diff --git a/evidence/indexer-255-blind-cue-20261030/README.md b/evidence/indexer-255-blind-cue-20261030/README.md new file mode 100644 index 000000000..134134dbe --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/README.md @@ -0,0 +1,5 @@ +# Blind task cue preparation: INCONCLUSIVE + +[Source-label decision](results/source-label-review.json):3 supported pairs below the frozen4-case minimum. Six authored task/cue pairs passed intent review, but no retrieval evaluation was run. Do not replace unsupported tasks or infer raw/cue effectiveness. Production baseline b3549ce1. + +Inputs/protocol are in plan/; review receipts in results/. accepted-receipt.json and blind-author.json are explicitly publication-redacted copies: only private local identifiers/paths were replaced. plan/publication-redactions.json records original and published SHA256. Frozen downstream input hashes and blind-author.json.sha256 intentionally retain original provenance; they are NOT asserted as the hashes of the redacted public copy. No task/cue wording or outcome was changed. Original copies remain locally retained. No runtime/candidate patch exists because evaluation never started; patch/README.md records that boundary. diff --git a/evidence/indexer-255-blind-cue-20261030/patch/README.md b/evidence/indexer-255-blind-cue-20261030/patch/README.md new file mode 100644 index 000000000..0da0bfbfe --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/patch/README.md @@ -0,0 +1 @@ +No candidate or runtime changes. Preparation stopped INCONCLUSIVE before paired retrieval. diff --git a/evidence/indexer-255-blind-cue-20261030/plan/accepted-receipt.json b/evidence/indexer-255-blind-cue-20261030/plan/accepted-receipt.json new file mode 100644 index 000000000..5d6c39e03 --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/plan/accepted-receipt.json @@ -0,0 +1,10 @@ +{ + "status": "COORDINATOR_ACCEPTED_BEFORE_CASE_AUTHORING_EVALUATION", + "accepted_at_utc": "2026-10-30T03:33:30.132560+00:00", + "protocol_sha256": "0dbe4843e50fc064934e98cad369024ae239b36c69772e2071d7e96c4eecbf17", + "baseline": "b3549ce1aea1556f76de1d2e14a4240474947b6b", + "scope": "One fixed paired raw-task vs manually extracted verbatim cue study; production scorer unchanged", + "author": "", + "next_gate": "Freeze six raw/cue pairs; independent intent review before source labelling; fewer than4 source-supported pairs stops INCONCLUSIVE without replacements", + "heldout": "Old six remain unused; no access needed" +} diff --git a/evidence/indexer-255-blind-cue-20261030/plan/blind-author.json b/evidence/indexer-255-blind-cue-20261030/plan/blind-author.json new file mode 100644 index 000000000..2b2df8169 --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/plan/blind-author.json @@ -0,0 +1,423 @@ +{ + "schema": "blind-manual-cue-authoring-v1", + "frozen_at_utc": "2026-10-30T03:34:44.915639+00:00", + "protocol_path": "/protocol-draft.md", + "protocol_sha256": "0dbe4843e50fc064934e98cad369024ae239b36c69772e2071d7e96c4eecbf17", + "author_access_attestation": { + "inputs_read": [ + "Assigned neutral engineering-task remit", + "protocol-draft.md", + "Intent reviewer coordination messages without catalog content" + ], + "not_accessed": [ + "InferRSI repository or catalog", + "skill titles", + "previous cases or results", + "source labels", + "scorer or search output", + "model service" + ], + "basis": "Self-reported process history, not forensic access isolation. General engineering knowledge used." + }, + "editing_rule": "Manually select verbatim OPERATION, OBJECT, CONSTRAINTS spans, then deterministically serialize them with semicolon-space. This is not an automatic semantic extractor. No synonyms, added keywords, answer hints, translations or catalog-informed edits.", + "offset_convention": "Zero-based Unicode code-point offsets, end-exclusive. All task characters happen to be ASCII. Removed spans partition all unselected raw text, including original separators; separators are reintroduced by serialization.", + "token_count_method": "Whitespace-delimited tokens via Python str.split(); descriptive counts only, not model or production-ranking tokenizer counts.", + "effort": { + "started_at_utc": "2026-10-30T03:33:14+00:00", + "frozen_at_utc": "2026-10-30T03:34:44.915639+00:00", + "elapsed_seconds": 90.915639, + "process": "One fresh authoring pass for six tasks followed by programmatic offset, coverage, serialization and identity assertions. No retrieval, catalog searches, scoring or outcome-driven revisions." + }, + "cases": [ + { + "case_id": "fresh255-01", + "intent": "Inference latency diagnosis", + "raw_task": "Could you help me with this? Diagnose whether queueing rather than model execution explains the high tail latency. Our text-generation endpoint serves short interactive requests alongside long background requests on one GPU. Keep the existing model and output quality unchanged, do not assume batching is the cause, and propose measurements that distinguish waiting time from execution time before recommending a configuration change.", + "cue": "Diagnose whether queueing rather than model execution explains the high tail latency.; Our text-generation endpoint serves short interactive requests alongside long background requests on one GPU.; Keep the existing model and output quality unchanged, do not assume batching is the cause, and propose measurements that distinguish waiting time from execution time before recommending a configuration change.", + "identity_control": false, + "spans": [ + { + "slot": "OPERATION", + "start": 29, + "end": 114, + "text": "Diagnose whether queueing rather than model execution explains the high tail latency." + }, + { + "slot": "OBJECT", + "start": 115, + "end": 224, + "text": "Our text-generation endpoint serves short interactive requests alongside long background requests on one GPU." + }, + { + "slot": "CONSTRAINTS", + "start": 225, + "end": 434, + "text": "Keep the existing model and output quality unchanged, do not assume batching is the cause, and propose measurements that distinguish waiting time from execution time before recommending a configuration change." + } + ], + "removed_spans": [ + { + "start": 0, + "end": 29, + "text": "Could you help me with this? ", + "reason": "politeness", + "detail": "Polite request opening." + }, + { + "start": 114, + "end": 115, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + }, + { + "start": 224, + "end": 225, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + } + ], + "constraints_retained": [ + "Keep the existing model and output quality unchanged, do not assume batching is the cause, and propose measurements that distinguish waiting time from execution time before recommending a configuration change." + ], + "operation_and_object_retained": [ + "Diagnose whether queueing rather than model execution explains the high tail latency.", + "Our text-generation endpoint serves short interactive requests alongside long background requests on one GPU." + ], + "counts": { + "raw_characters": 434, + "cue_characters": 407, + "raw_tokens": 62, + "cue_tokens": 56 + }, + "compression_note": "No semantic information removed. Only politeness and inter-sentence whitespace removed/replaced; sentence punctuation retained verbatim.", + "raw_sha256": "d9bcf58de3a254646c12e0ec8ec9b80e7b4294a26e8e051c31ca63a3c754a116", + "cue_sha256": "47a391fbe286c5a3a702dff48c57b5542d76b094f4a61430de67b3f111588f4f" + }, + { + "case_id": "fresh255-02", + "intent": "Distributed training hang diagnosis", + "raw_task": "Identify likely causes of a distributed training hang and give a minimal diagnostic sequence; PyTorch DistributedDataParallel on four GPUs with gradient accumulation; do not disable gradient synchronization as a fix; preserve the effective batch size", + "cue": "Identify likely causes of a distributed training hang and give a minimal diagnostic sequence; PyTorch DistributedDataParallel on four GPUs with gradient accumulation; do not disable gradient synchronization as a fix; preserve the effective batch size", + "identity_control": true, + "spans": [ + { + "slot": "OPERATION", + "start": 0, + "end": 92, + "text": "Identify likely causes of a distributed training hang and give a minimal diagnostic sequence" + }, + { + "slot": "OBJECT", + "start": 94, + "end": 165, + "text": "PyTorch DistributedDataParallel on four GPUs with gradient accumulation" + }, + { + "slot": "CONSTRAINTS", + "start": 167, + "end": 250, + "text": "do not disable gradient synchronization as a fix; preserve the effective batch size" + } + ], + "removed_spans": [ + { + "start": 92, + "end": 94, + "text": "; ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + }, + { + "start": 165, + "end": 167, + "text": "; ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + } + ], + "constraints_retained": [ + "do not disable gradient synchronization as a fix; preserve the effective batch size" + ], + "operation_and_object_retained": [ + "Identify likely causes of a distributed training hang and give a minimal diagnostic sequence", + "PyTorch DistributedDataParallel on four GPUs with gradient accumulation" + ], + "counts": { + "raw_characters": 250, + "cue_characters": 250, + "raw_tokens": 35, + "cue_tokens": 35 + }, + "compression_note": "No semantic information removed. Byte-identical control; original slot delimiters are reinserted unchanged.", + "raw_sha256": "ef6d48dde1e99286b048e6d19301041f50824813b18c5c240647b7e8b1c4926b", + "cue_sha256": "ef6d48dde1e99286b048e6d19301041f50824813b18c5c240647b7e8b1c4926b" + }, + { + "case_id": "fresh255-03", + "intent": "Export dynamic-shape compatibility", + "raw_task": "Please help me work through the following. Explain how to export this model and verify that the exported program accepts the intended input shapes. The target API is torch.export.export, and the model takes token IDs shaped [batch, sequence] plus an attention mask with the same shape. Batch size must remain fixed at 1, sequence length must vary from 16 through 512 inclusive, invalid shapes must be rejected, and the answer must include a small validation example rather than switching to a different export API.", + "cue": "Explain how to export this model and verify that the exported program accepts the intended input shapes.; The target API is torch.export.export, and the model takes token IDs shaped [batch, sequence] plus an attention mask with the same shape.; Batch size must remain fixed at 1, sequence length must vary from 16 through 512 inclusive, invalid shapes must be rejected, and the answer must include a small validation example rather than switching to a different export API.", + "identity_control": false, + "spans": [ + { + "slot": "OPERATION", + "start": 43, + "end": 147, + "text": "Explain how to export this model and verify that the exported program accepts the intended input shapes." + }, + { + "slot": "OBJECT", + "start": 148, + "end": 285, + "text": "The target API is torch.export.export, and the model takes token IDs shaped [batch, sequence] plus an attention mask with the same shape." + }, + { + "slot": "CONSTRAINTS", + "start": 286, + "end": 514, + "text": "Batch size must remain fixed at 1, sequence length must vary from 16 through 512 inclusive, invalid shapes must be rejected, and the answer must include a small validation example rather than switching to a different export API." + } + ], + "removed_spans": [ + { + "start": 0, + "end": 43, + "text": "Please help me work through the following. ", + "reason": "politeness", + "detail": "Polite request opening." + }, + { + "start": 147, + "end": 148, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + }, + { + "start": 285, + "end": 286, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + } + ], + "constraints_retained": [ + "Batch size must remain fixed at 1, sequence length must vary from 16 through 512 inclusive, invalid shapes must be rejected, and the answer must include a small validation example rather than switching to a different export API." + ], + "operation_and_object_retained": [ + "Explain how to export this model and verify that the exported program accepts the intended input shapes.", + "The target API is torch.export.export, and the model takes token IDs shaped [batch, sequence] plus an attention mask with the same shape." + ], + "counts": { + "raw_characters": 514, + "cue_characters": 473, + "raw_tokens": 84, + "cue_tokens": 77 + }, + "compression_note": "No semantic information removed. Only politeness and inter-sentence whitespace removed/replaced; sentence punctuation retained verbatim.", + "raw_sha256": "525c6f2067999ec6f1855e3e77752a33597b75dc3b2aa4d87220210447b46c87", + "cue_sha256": "67db4e34837f94c68cecb778214f4c27878de86faefac0efebda186d7152ec4a" + }, + { + "case_id": "fresh255-04", + "intent": "Training checkpoint recovery correctness", + "raw_task": "Plan a checkpoint-and-resume test; a single-GPU training run using AdamW, dropout, and shuffled minibatches; restore optimizer and random-number-generator state; compare the next ten losses with an uninterrupted run; do not claim bitwise reproducibility across different hardware", + "cue": "Plan a checkpoint-and-resume test; a single-GPU training run using AdamW, dropout, and shuffled minibatches; restore optimizer and random-number-generator state; compare the next ten losses with an uninterrupted run; do not claim bitwise reproducibility across different hardware", + "identity_control": true, + "spans": [ + { + "slot": "OPERATION", + "start": 0, + "end": 33, + "text": "Plan a checkpoint-and-resume test" + }, + { + "slot": "OBJECT", + "start": 35, + "end": 107, + "text": "a single-GPU training run using AdamW, dropout, and shuffled minibatches" + }, + { + "slot": "CONSTRAINTS", + "start": 109, + "end": 279, + "text": "restore optimizer and random-number-generator state; compare the next ten losses with an uninterrupted run; do not claim bitwise reproducibility across different hardware" + } + ], + "removed_spans": [ + { + "start": 33, + "end": 35, + "text": "; ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + }, + { + "start": 107, + "end": 109, + "text": "; ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + } + ], + "constraints_retained": [ + "restore optimizer and random-number-generator state; compare the next ten losses with an uninterrupted run; do not claim bitwise reproducibility across different hardware" + ], + "operation_and_object_retained": [ + "Plan a checkpoint-and-resume test", + "a single-GPU training run using AdamW, dropout, and shuffled minibatches" + ], + "counts": { + "raw_characters": 279, + "cue_characters": 279, + "raw_tokens": 36, + "cue_tokens": 36 + }, + "compression_note": "No semantic information removed. Byte-identical control; original slot delimiters are reinserted unchanged.", + "raw_sha256": "8267376b490315099ac8e1ed4aa2cf285abc19200ec614aa2c04457a075febc0", + "cue_sha256": "8267376b490315099ac8e1ed4aa2cf285abc19200ec614aa2c04457a075febc0" + }, + { + "case_id": "fresh255-05", + "intent": "Inference memory admission design", + "raw_task": "I would appreciate your help. Design an admission policy that prevents out-of-memory failures without silently dropping accepted requests. Our image-and-text inference service receives images with different resolutions and prompts with different token lengths, and several requests can be active on the same GPU. Account for input-dependent memory rather than using request count alone, state how to handle a request that cannot fit even when run by itself, and describe how to test the policy without requiring production traffic. Thanks.", + "cue": "Design an admission policy that prevents out-of-memory failures without silently dropping accepted requests.; Our image-and-text inference service receives images with different resolutions and prompts with different token lengths, and several requests can be active on the same GPU.; Account for input-dependent memory rather than using request count alone, state how to handle a request that cannot fit even when run by itself, and describe how to test the policy without requiring production traffic.", + "identity_control": false, + "spans": [ + { + "slot": "OPERATION", + "start": 30, + "end": 138, + "text": "Design an admission policy that prevents out-of-memory failures without silently dropping accepted requests." + }, + { + "slot": "OBJECT", + "start": 139, + "end": 312, + "text": "Our image-and-text inference service receives images with different resolutions and prompts with different token lengths, and several requests can be active on the same GPU." + }, + { + "slot": "CONSTRAINTS", + "start": 313, + "end": 531, + "text": "Account for input-dependent memory rather than using request count alone, state how to handle a request that cannot fit even when run by itself, and describe how to test the policy without requiring production traffic." + } + ], + "removed_spans": [ + { + "start": 0, + "end": 30, + "text": "I would appreciate your help. ", + "reason": "politeness", + "detail": "Polite request opening." + }, + { + "start": 138, + "end": 139, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + }, + { + "start": 312, + "end": 313, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + }, + { + "start": 531, + "end": 539, + "text": " Thanks.", + "reason": "politeness", + "detail": "Polite closing." + } + ], + "constraints_retained": [ + "Account for input-dependent memory rather than using request count alone, state how to handle a request that cannot fit even when run by itself, and describe how to test the policy without requiring production traffic." + ], + "operation_and_object_retained": [ + "Design an admission policy that prevents out-of-memory failures without silently dropping accepted requests.", + "Our image-and-text inference service receives images with different resolutions and prompts with different token lengths, and several requests can be active on the same GPU." + ], + "counts": { + "raw_characters": 539, + "cue_characters": 503, + "raw_tokens": 79, + "cue_tokens": 73 + }, + "compression_note": "No semantic information removed. Only politeness and inter-sentence whitespace removed/replaced; sentence punctuation retained verbatim.", + "raw_sha256": "ee36503d3c192f10b6c85be111a0bea1445aef0c287f6bf1a34db903dc1b77e0", + "cue_sha256": "a34321b804aa1a57ad8da828b32478c4e9b75dcafbd1e05f2f56f2d9f51d21d7" + }, + { + "case_id": "fresh255-06", + "intent": "Attention numerical discrepancy investigation", + "raw_task": "Could you take a look at this? Investigate whether the observed output differences are expected numerical error or an implementation defect. A causal attention implementation in half precision disagrees with a full-precision reference most strongly on long sequences with padding. Check masking and padded positions as well as arithmetic precision, do not assume the optimized kernel is wrong, and provide a small reproducible comparison with explicit absolute and relative error measurements before suggesting a replacement.", + "cue": "Investigate whether the observed output differences are expected numerical error or an implementation defect.; A causal attention implementation in half precision disagrees with a full-precision reference most strongly on long sequences with padding.; Check masking and padded positions as well as arithmetic precision, do not assume the optimized kernel is wrong, and provide a small reproducible comparison with explicit absolute and relative error measurements before suggesting a replacement.", + "identity_control": false, + "spans": [ + { + "slot": "OPERATION", + "start": 31, + "end": 140, + "text": "Investigate whether the observed output differences are expected numerical error or an implementation defect." + }, + { + "slot": "OBJECT", + "start": 141, + "end": 280, + "text": "A causal attention implementation in half precision disagrees with a full-precision reference most strongly on long sequences with padding." + }, + { + "slot": "CONSTRAINTS", + "start": 281, + "end": 525, + "text": "Check masking and padded positions as well as arithmetic precision, do not assume the optimized kernel is wrong, and provide a small reproducible comparison with explicit absolute and relative error measurements before suggesting a replacement." + } + ], + "removed_spans": [ + { + "start": 0, + "end": 31, + "text": "Could you take a look at this? ", + "reason": "politeness", + "detail": "Polite request opening." + }, + { + "start": 140, + "end": 141, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + }, + { + "start": 280, + "end": 281, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot delimiter is replaced by the mandated serialization delimiter; no semantic content is removed." + } + ], + "constraints_retained": [ + "Check masking and padded positions as well as arithmetic precision, do not assume the optimized kernel is wrong, and provide a small reproducible comparison with explicit absolute and relative error measurements before suggesting a replacement." + ], + "operation_and_object_retained": [ + "Investigate whether the observed output differences are expected numerical error or an implementation defect.", + "A causal attention implementation in half precision disagrees with a full-precision reference most strongly on long sequences with padding." + ], + "counts": { + "raw_characters": 525, + "cue_characters": 496, + "raw_tokens": 75, + "cue_tokens": 68 + }, + "compression_note": "No semantic information removed. Only politeness and inter-sentence whitespace removed/replaced; sentence punctuation retained verbatim.", + "raw_sha256": "4d386903bdb05c32dcb853128e531ad9e898d9ea8ea46a2c0c947311f4cd98bb", + "cue_sha256": "193bb70f0cd2ef507139f621c78f4f122fc287aafcb5adfcd44ba7e0b0761163" + } + ], + "gate_status": "Input author frozen. Independent intent review pending; source classification and retrieval not performed." +} diff --git a/evidence/indexer-255-blind-cue-20261030/plan/blind-author.json.sha256 b/evidence/indexer-255-blind-cue-20261030/plan/blind-author.json.sha256 new file mode 100644 index 000000000..d1fc134f7 --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/plan/blind-author.json.sha256 @@ -0,0 +1 @@ +ed81b75333efaab1d8834ff6f9d30f2b060b070212ee463d9b7395e0a0cfcfc3 blind-author.json diff --git a/evidence/indexer-255-blind-cue-20261030/plan/protocol-draft.md b/evidence/indexer-255-blind-cue-20261030/plan/protocol-draft.md new file mode 100644 index 000000000..cdfc3dc2a --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/plan/protocol-draft.md @@ -0,0 +1,47 @@ +# #255 input-expression feasibility study — prospective draft + +Status: DRAFT FOR COORDINATOR ACCEPTANCE before case authoring or retrieval. Owner accepted this direction at03:30 UTC2026-10-30; no scorer change or automatic query rewrite was authorized by this study. One bounded CPU comparison, six fresh development tasks. No paid/model/hardware calls, installs, external writes or old-heldout access. + +## Question and baseline +Can task-preserving, manually prepared `操作＋对象＋约束` relevance cues improve retrieval ordering compared with the same task expressed as a longer narrative, using unchanged production ranking? + +Pin production `b3549ce1aea1556f76de1d2e14a4240474947b6b` and the existing813-file source manifest SHA256 `9a1a7d9fda980602ca344a4cf8be10fbfa7eef96c2dc1feaea8eab75647c4215`. Refresh main before acceptance; if source/corpus changed, pin and disclose the new baseline before inputs rather than quietly comparing across snapshots. Runtime/source and corpus must be identical for both arms. Existing weighted-lexical-v1 only; no v2/v3 opt-in mode, code/scorer changes, field weights, filters, query tuning or metadata edits. + +## Roles and information barriers +1. A fresh input author with no catalog, target-label, old case wording or retrieval-result access authors exactly six plausible engineering tasks and concise cues. Supply only this protocol plus a broad neutral remit: inference/training engineering tasks involving a concrete operation, object and constraint; do not supply selected skill names, keyword lists, old failure examples or intended targets. Task author records an access attestation and cue-editing time/effort. +2. An independent intent reviewer checks each raw/cue pair BEFORE source labelling: operations/objects/constraints, negation, causal uncertainty, environment, thresholds and requested deliverable must agree. Reviewer does not rewrite cues or inject repository terminology. Disagreement marks the pair invalid; no replacement after seeing retrieval results. +3. A separate source-label reviewer may inspect the frozen six tasks and pinned corpus, then assigns partial positive IDs and exact source support. No rank calls, scorer simulation or candidate outputs. Do not change task/cue wording after seeing catalog matches. Record unlabelled/ambiguous cases honestly, not forced targets. +4. Evaluator receives the final input commitment only after all above receipts. One paired retrieval run; report both arms including misses. Old six sealed heldout cases remain sealed and are neither accessed nor repurposed. + +A role-separated author process can use subagents, but these are human/assistant-curated experimental inputs, NOT observed user traffic or an automatic normalizer. Attestations/history establish the intended barrier; do not claim forensic access isolation. + +## Fixed cue-editing rule (manual, auditable, no model rewrite service) +The author first writes the raw task as a natural narrative, then annotates verbatim spans into three ordered slots: + +- OPERATION: the actual action/question to answer, including uncertainty or requested diagnosis rather than turning it into a prescription. +- OBJECT: workload/component/data object and explicit environment identifiers needed to interpret it. +- CONSTRAINTS: ALL explicit correctness/safety/resource/performance conditions, negations, thresholds and required output. + +Construct the cue by joining those verbatim spans in order with `; `. Do not paraphrase, add synonyms/keywords, singularize, translate, remove negation, introduce desired answers or rename objects after catalog terms. Repeated exact spans may appear once; otherwise preserve the selected text. Remove only greetings, politeness, redundant framing or incidental chronology not used by the task. If meaningful information cannot be removed safely, retain it; a cue need not be short. Record every removed span and a reason from this fixed list, plus raw/cue character and token counts. This is a deterministic serialization of manually selected slots, NOT deterministic automatic semantic extraction. Report that distinction explicitly. + +To prevent an artificial advantage, raw tasks must not be padded with arbitrary unrelated vocabulary, intentionally bait substring matches, or hide a known skill title. At least two naturally concise tasks should act as identity/near-identity controls: if they already consist of operation/object/constraints, keep the wording unchanged. Do not search for tasks that produce a requested score. + +## Input counts and pre-evaluation validity gate +Exactly six fresh tasks, at least four distinct task intents, including at least two identity/near-identity controls within six. Distinctness reviewed semantically, not proven statistical independence. Avoid reusing previous case wording or known failed tasks; the coordinator can check overlap using IDs/hashes and coarse intent without disclosing old sealed labels to author. + +All six pairs must pass intent preservation. At least four of six must have source-supported positives; otherwise preparation yields INCONCLUSIVE coverage and STOP before retrieval. Freeze all six original pairs, all partial labels/ambiguities, source hashes, retained/removed-span maps and the eligible metric denominator. Do not replace source-uncovered cases just to manufacture success. Remaining unlabelled cases still run and are reported descriptively, not called failures or true negatives. No labels may be added after viewing output ranks for the gate calculation. + +## Retrieval settings and prospective gates +Both arms use `query=''`, no filters, `limit=5`, `context_bytes=262144`, same production `relevance` interface. Only the cue string differs. Paired outputs captured once per frozen pair; no retries/tuning/relabel after outcomes. + +- All pre-evaluation validity checks pass. +- On frozen source-labelled eligible pairs, no baseline positive Hit@5 is lost; Hit@5 is nondecreasing; MRR@5 is STRICTLY greater than raw-task baseline. Minimum benefit is ordinal/effect direction, not statistical confidence. Report absolute per-case rank changes, denominator, summed RR and MRR; no rounding into a pass. +- Identity controls yield exact serialized output parity when raw and cue are byte-identical. Near-identity controls remain descriptive pairs, not an exact-parity assertion. At least one byte-identical control is required. +- Existing six unchanged/default contract captures have exact parity against baseline; source/scorer hash equality and no code diff mandatory. +- Full indexed_skill mandatory resources, emitted bytes/hashes/revision, context-budget accounting and explicit omissions must pass in both arms; complete relevant delivery count must not decrease. Required suffixes alone are insufficient. +- Any valid negative gate result is FAIL. Missing/broken input identity or too few source-supported cases is INCONCLUSIVE. Stop; retain all outputs and no second rule/case replacement/search-until-pass. + +## Decision and limits +PASS supports only feasibility of this manually prepared input contract on a small curated set. It does not prove an automatic rewriting algorithm, lower user effort, speed, population accuracy, complete precision/recall, downstream agent benefit or hardware efficacy. Require reporting how much task-author effort was needed and any information that could not be compressed. No shipping caller burden, default change, promotion or #255 closure follows automatically. A next validation/implementation decision is separate. + +Previous cache/v2/v3 FAIL reports remain unchanged. The six old heldout cases stay unused regardless of this study's outcome. diff --git a/evidence/indexer-255-blind-cue-20261030/plan/publication-integrity.json b/evidence/indexer-255-blind-cue-20261030/plan/publication-integrity.json new file mode 100644 index 000000000..c7a37481c --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/plan/publication-integrity.json @@ -0,0 +1,12 @@ +{ + "README.md": "b44e6518244cb4fb51588a1ed080d363b28ca267e6c7363b939772349e728a9e", + "patch/README.md": "33a32a2d7dc0c4f14122e192bbcd6b4e58630aac858852054103e8f58afa6121", + "plan/accepted-receipt.json": "4b90d0d1b2cacd16e822f6cff78ea7532d1fc915a2ca366c00a1cad5f2f5d94f", + "plan/blind-author.json": "66bb2a2c47e18504d6e0650fbc4f06cc9e785e71f44840ef985e65e8e0e9e54c", + "plan/blind-author.json.sha256": "e36e17adb28b0cdd5eb863b706b39b6886f726164bc71348e44a4cee0b095e56", + "plan/protocol-draft.md": "0dbe4843e50fc064934e98cad369024ae239b36c69772e2071d7e96c4eecbf17", + "plan/publication-redactions.json": "7a5a859dea495dc2f52baf1de07b805370bf1b749a76f470885165783d2d31f3", + "results/intent-review.json": "1aa230da216f6d5ad918cd69dbf181cfec0525b4168f7dad72d0dc1d4304a627", + "results/source-label-review.json": "3ca52b5576fca03c7e536d6c19b4ff2ff2b0c88c6c8a63dfcdfdf2d5de5f586f", + "validator/README.md": "7a3ef259da87821a93308b026ee33b31eb3e828b0bf2d5443fad8e7d2b816cbc" +} diff --git a/evidence/indexer-255-blind-cue-20261030/plan/publication-redactions.json b/evidence/indexer-255-blind-cue-20261030/plan/publication-redactions.json new file mode 100644 index 000000000..de2f53caa --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/plan/publication-redactions.json @@ -0,0 +1,14 @@ +[ + { + "artifact": "plan/blind-author.json", + "original_sha256": "ed81b75333efaab1d8834ff6f9d30f2b060b070212ee463d9b7395e0a0cfcfc3", + "published_sha256": "66bb2a2c47e18504d6e0650fbc4f06cc9e785e71f44840ef985e65e8e0e9e54c", + "change": "Only local task identifier or absolute scratch path replaced by labelled placeholder. Frozen task/cue strings unchanged; original hashes in receipts refer to retained local original." + }, + { + "artifact": "plan/accepted-receipt.json", + "original_sha256": "7d99ed6d16846a788d6ac2410b5b93548814a83b318b373d91912d5ca4cd1121", + "published_sha256": "4b90d0d1b2cacd16e822f6cff78ea7532d1fc915a2ca366c00a1cad5f2f5d94f", + "change": "Only local task identifier or absolute scratch path replaced by labelled placeholder. Frozen task/cue strings unchanged; original hashes in receipts refer to retained local original." + } +] diff --git a/evidence/indexer-255-blind-cue-20261030/results/intent-review.json b/evidence/indexer-255-blind-cue-20261030/results/intent-review.json new file mode 100644 index 000000000..50c33efd8 --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/results/intent-review.json @@ -0,0 +1,48 @@ +{ + "reviewed_at_utc": "2026-10-30T03:35:10.365123+00:00", + "input_sha256": "ed81b75333efaab1d8834ff6f9d30f2b060b070212ee463d9b7395e0a0cfcfc3", + "status": "PASS_INTENT_PRESERVATION_BEFORE_SOURCE_LABEL_REVIEW", + "cases": [ + { + "id": "fresh255-01", + "verbatim_spans_valid": true, + "all_explicit_constraints_and_uncertainty_retained": true, + "identity_control": false + }, + { + "id": "fresh255-02", + "verbatim_spans_valid": true, + "all_explicit_constraints_and_uncertainty_retained": true, + "identity_control": true + }, + { + "id": "fresh255-03", + "verbatim_spans_valid": true, + "all_explicit_constraints_and_uncertainty_retained": true, + "identity_control": false + }, + { + "id": "fresh255-04", + "verbatim_spans_valid": true, + "all_explicit_constraints_and_uncertainty_retained": true, + "identity_control": true + }, + { + "id": "fresh255-05", + "verbatim_spans_valid": true, + "all_explicit_constraints_and_uncertainty_retained": true, + "identity_control": false + }, + { + "id": "fresh255-06", + "verbatim_spans_valid": true, + "all_explicit_constraints_and_uncertainty_retained": true, + "identity_control": false + } + ], + "limits": [ + "Reviewer has prior general repository knowledge but has not accessed source/catalog for these new pairs before this intent receipt. Author independently attests no catalog/old-case/rank access.", + "Four nonidentity cases remove only polite framing, not substantive narrative. This is a narrower politeness-removal/slot-serialization test, not demonstrated semantic compression. Do not generalize outcome to arbitrary normalization.", + "No query edits requested. Source labels next; preserve exact input hash." + ] +} diff --git a/evidence/indexer-255-blind-cue-20261030/results/source-label-review.json b/evidence/indexer-255-blind-cue-20261030/results/source-label-review.json new file mode 100644 index 000000000..f6d60b47e --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/results/source-label-review.json @@ -0,0 +1,114 @@ +{ + "reviewed_at_utc": "2026-10-30T03:36:39.961469+00:00", + "baseline": "b3549ce1aea1556f76de1d2e14a4240474947b6b", + "input_sha256": "ed81b75333efaab1d8834ff6f9d30f2b060b070212ee463d9b7395e0a0cfcfc3", + "intent_review_sha256": "1aa230da216f6d5ad918cd69dbf181cfec0525b4168f7dad72d0dc1d4304a627", + "method": "Catalog inventory and bounded source-text inspection/search only; no scorer, collect, ranking simulation or retrieval evaluation. Queries/labels never edited based on outputs.", + "cases": [ + { + "id": "fresh255-01", + "status": "SOURCE_SUPPORTED_PARTIAL_POSITIVE", + "relevant_ids": [ + "inferrsi.common.skills.assess_serving_preemption" + ], + "evidence": [ + { + "path": "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md", + "sha256": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d", + "start_line": 12, + "end_line": 25, + "quote": "## Distinguish the failed boundary\n\nEstablish what is late: first token, individual token gap, request completion, or a stated cohort's service. Obtain the exact percentile/attainment target and denominator. Then separate:\n- Waiting for the current whole batch to end\n- Inability to admit because reserved or live KV is full\n- Long first-iteration/prefill interference\n- Long accepted requests occupying service while short jobs queue\n- Waiting for state recovery or host/device transfer\n\nIteration-level batching can address the first without solving the others. Short-job preemption addresses service ordering but can worsen the total unfinished-state footprint. If prefill interference is the binding problem, the optional earlier `choose-prefill-decode-schedule` candidate addresses that separate question; `diagnose-kv-block-efficiency` addresses physical allocation/slack. Neither is required to consume this skill: if unavailable, report the diagnosed boundary and the missing current evidence rather than substituting a preemption recipe.\n\n## Evidence required for a candidate\n\nRecord model/KV format, actual engine revision, safe interruption boundary, request batch limit, input/output/context distributions, arrival bursts, device live/reserved/allocated bytes, active and suspended requests, host budget, transfer path, recovery time, and queue/phase/service timings. Keep per-tenant or priority-class obligations distinct from aggregate latency. Mark unavailable evidence explicitly." + } + ], + "limits": "Queue delay/service ordering and discriminating timing evidence directly relevant; not a complete diagnosis or current engine qualification." + }, + { + "id": "fresh255-02", + "status": "UNLABELLED_SCOPE_MISMATCH", + "relevant_ids": [], + "evidence": [ + { + "path": "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md", + "sha256": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f", + "start_line": 20, + "end_line": 25, + "quote": "\nRuntime configuration is published into namespace bags; the raw ServerArgs record is not the effective decision source. Inspect the current native runtime-context contract before modifying config reads, overrides, or per-runner state.\n\nExpected result: topology/ownership diagram or precise patch boundary plus baseline comparisons. Validate one rank first when appropriate, then the intended groups and PD peers. Include idle ranks, uneven loads, cancellation, and error propagation in distributed checks.\n\nA distributed hang requires rank-specific evidence and collective-order inspection. A numerically wrong kernel requires a reduced input and reference comparison. Revert the affected backend/feature to the baseline; do not infer correctness from a successful import, launch, or faster microbenchmark. Plan the [validation gates](../../references/tuning-and-validation.md) before accepting the extension." + } + ], + "limits": "Nearby rank-specific collective guidance is SGLang extension validation, not PyTorch DistributedDataParallel training with gradient accumulation/effective-batch preservation. No sufficiently specific positive established; not proof no useful resource exists." + }, + { + "id": "fresh255-03", + "status": "UNLABELLED_NO_SPECIFIC_SOURCE_ESTABLISHED", + "relevant_ids": [], + "evidence": [], + "limits": "No specific torch.export.export dynamic-shape constraint/validation guidance established in bounded skill-library review. Do not substitute another export API or generic model adaptation." + }, + { + "id": "fresh255-04", + "status": "SOURCE_SUPPORTED_PARTIAL_POSITIVE", + "relevant_ids": [ + "inferrsi.common.skills.checkpoint_recovery_consistency" + ], + "evidence": [ + { + "path": "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md", + "sha256": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688", + "start_line": 25, + "end_line": 35, + "quote": "## Establish four independent acceptance questions\n\n| Question | Required evidence |\n|---|---|\n| Does the snapshot survive the declared failure? | Actual memory/storage/replica placement and helper lifetime |\n| Is one generation complete across required participants? | Step/iteration identity, expected shard/rank set, successful finalization |\n| Does that generation contain the application's complete state? | Model, optimizer and caller-required RNG/scheduler/data-position state contract |\n| Can the target environment restore it? | Matching format, world-size/replication assumptions and separately authorized restore evidence |\n\nNeither equal steps nor complete rank IDs verifies tensor contents or whole\napplication semantics. Record these as distinct evidence levels." + } + ], + "limits": "Direct model/optimizer/RNG/scheduler/data-position completeness contract useful for resume test; does not supply AdamW/dropout code or prove next-ten-loss or cross-hardware parity. Partial relevance, not complete task solution." + }, + { + "id": "fresh255-05", + "status": "UNLABELLED_SCOPE_MISMATCH", + "relevant_ids": [], + "evidence": [ + { + "path": "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md", + "sha256": "d99fc28d48127d3bc627f2a0cab6f3f16158a77da49428838a08865adaea9905", + "start_line": 40, + "end_line": 51, + "quote": "## Required evidence\n\nCollect the deployed revision, scheduler class, effective scheduler/cache configuration, connector class and role, KV-group layout, scheduling policy, speculative/encoder features, and concurrent-batch mode. Obtain a bounded request/step trace: queue/status, actual scheduled tokens, both remaining budgets, allocation result, preemption victim, pending-transfer/free state, and request timing. Separate unavailable observations from zero values. `Scheduler.make_stats` is gated by `log_stats`; its queue/KV fields are useful context but do not expose every branch reason (F7).\n\n## Decision procedure\n\n1. Establish the effective caps: running slots derive from `max_num_seqs`; admission uses `max_num_active_seqs` or its fallback; scheduling-token and input budgets are separate (F1). Record both, including draft-slot charges. Never treat a nonempty queue as proof of memory exhaustion.\n2. Trace the first blocking/skip branch for representative requests. Running requests are considered first. Token availability, context bounds, cadence, encoder constraints, and alignment can prevent work before KV allocation (F2). Distinguish a skipped request from a loop-ending failure.\n3. For a long prefill, check whether more than one request is eligible before attributing a cap to `long_prefill_token_threshold`; the singleton case disables it and the adaptive option can raise the cap (F2). Do not copy a default or prescribe a universal threshold.\n4. If `allocate_slots` returns `None`, inspect `has_pending_block_frees` and `_request_blocks_can_be_freed` before blaming victim selection. The source can stop allocation without preempting. Otherwise identify the actual priority/FCFS victim and, for an already-scheduled priority victim, restored token/input/encoder accounting (F3, F6).\n5. Reconstruct waiting admission separately: preemption earlier in the same step, pause state, KV-holding queue precedence, active-request cap, blocked status, and LoRA limit precede useful work (F4). For external-prefix lookup, distinguish unknown matches, zero matches, and positive asynchronous matches. Async admission reserves blocks while scheduling zero new tokens; its computed-token counter is not evidence of load completion (F5).\n6. Classify the observed limiting branch and one competing explanation. Produce one conditional hypothesis: adjust a documented cap only after checking the deployed configuration schema, or investigate connector progress instead. Predict which branch/metric should change and what latency, capacity, or recomputation trade-off could worsen. Do not change settings under this skill." + } + ], + "limits": "vLLM pinned scheduler branch diagnosis is related, but cannot be labelled a generic image-resolution/prompt-dependent admission-policy design with never-fit request handling and offline test. No target-runtime assumption added to raw task." + }, + { + "id": "fresh255-06", + "status": "SOURCE_SUPPORTED_PARTIAL_POSITIVE", + "relevant_ids": [ + "inferrsi.cuda.skills.diagnose_attention_bottlenecks" + ], + "evidence": [ + { + "path": "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md", + "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431", + "start_line": 12, + "end_line": 20, + "quote": "Use when attention remains slow despite fusion/tiling, a larger tile or deeper pipeline regresses, low-precision throughput disappoints, or a proposed hardware port assumes that the paper's instructions transfer directly.\n\nObtain the actual kernel/library revision and device; forward/backward or inference phase; batch, query/KV lengths, head count and head dimension; causal/padding masks, dropout and MQA/GQA; input, accumulator and intermediate dtypes; output/gradient tolerance and task-quality requirements. Distinguish an attention kernel from a whole training step or serving request.\n\nRequest existing evidence for HBM bytes and allocation, shared-memory/register use and spills, independent CTA/work-tile counts, synchronization/atomic waits, Tensor Core/GEMM activity, scalar/special-function work, and exposed load/compute stalls. A low percentage of peak FLOP/s does not identify which is limiting. If these observations are unavailable, return conditional hypotheses and the missing evidence; do not initiate profiling or runs without authorization in the current task.\n\n## Decision procedure\n\n1. **Check semantics before costs.** Preserve stable online-softmax normalization, masks and dropout replay; require correct output and gradients at the approved tolerance. Dense exact attention, block-sparse attention and quantized attention have different contracts. Exact algebra does not imply bitwise-identical floating-point results. Do not turn the historical papers' pseudocode directly into an implementation without resolving their notation/boundary issues. [F1\u2013F3; sources, interpretation cautions]" + }, + { + "path": "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md", + "sha256": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431", + "start_line": 35, + "end_line": 41, + "quote": "## Required output and proposed verification\n\nReturn: observed bottleneck and confidence; workload/semantic contract; one smallest candidate change; expected resource exchange; counter-signal; missing evidence; source ID; and the next authorization gate.\n\nFor a later authorized comparison, keep hardware/clock, stack, workload, dtype, mask/dropout and measurement boundary fixed. Report forward, backward and combined time separately; HBM bytes and allocation separately; fast-memory use/spills, active work and waits; preprocessing cost for precision changes; output/gradient error and any required task-quality result. State whether throughput counts useful matmuls, recomputation, causal masking and batch. Only claim application benefit after measuring that application boundary.\n\nReject or defer a proposal on wrong outputs/gradients, violated buffer lifetime, unverified compiler ordering, spills or memory overrun, unsupported primitives, unacceptable accuracy, or a gain that vanishes after including the omitted work. Empirical validation remains not_run; source-paper results do not establish local compatibility or efficacy." + } + ], + "limits": "Source requires mask/dtypes/tolerances and correctness before optimization, directly useful to numerical discrepancy investigation; does not implement the requested small comparison or universal tolerance." + } + ], + "source_supported_count": 3, + "required_minimum": 4, + "decision": "PREPARATION_INCONCLUSIVE_STOP_BEFORE_RETRIEVAL", + "heldout": "OLD_HELDOUT_UNTOUCHED", + "limits": [ + "Partial labels not exhaustive relevance truth or complete task coverage. Unlabelled is not irrelevant.", + "Do not replace tasks to force coverage or weaken4-case gate.", + "Intent reviewer had prior corpus knowledge; new author blind-access attestation and task input hash preserved.", + "No inference about raw-vs-cue effectiveness can be drawn because no paired retrieval was run." + ] +} diff --git a/evidence/indexer-255-blind-cue-20261030/validator/README.md b/evidence/indexer-255-blind-cue-20261030/validator/README.md new file mode 100644 index 000000000..2c042e1b1 --- /dev/null +++ b/evidence/indexer-255-blind-cue-20261030/validator/README.md @@ -0,0 +1 @@ +No retrieval validator was executed. Frozen preparation gates and source/intent reviews are the evidence; source-label-review.json records the stop before retrieval. diff --git a/evidence/indexer-255-covered-cue-20261030/README.md b/evidence/indexer-255-covered-cue-20261030/README.md new file mode 100644 index 000000000..702a21313 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/README.md @@ -0,0 +1,8 @@ +# Source-conditioned cue feasibility: neutral FAIL + +[Report](results/report.md):both arms Hit6/6, exactMRR29/36, complete relevant6/6; strictMRR improvement gate fails. Six source-conditioned synthetic cases, four mostly polite-framing edits and two byte-identical controls. Not broad normalization evidence. Earlier blind preparation-INCONCLUSIVE remains separate. + +Frozen production baseline6e6ee245 is identical in both arms. For authorized replay only, create a new output directory and run validator/evaluate.py BASELINE_ROOT PREP_ROOT OUTPUT_DIR with PREP_ROOT set to a new external copy of plan/ after the source-pins.json → source-manifest.json restoration described below. Do not overwrite frozen results. No runtime changes, scorer candidate, model/hardware experiment or #255 closure. Original local integrity inventory is replaced by plan/publication-integrity.json for relocation; frozen outputs/protocol/inputs/runners remain exact bytes. + +### Frozen preparation filename mapping +`plan/source-pins.json` is the exact original source-manifest.json snapshot (not a current-tree hash validator). For replay copy plan/ to a new external preparation directory and copy source-pins.json to source-manifest.json there; do not modify the evidence tree. Internal frozen references retain the original filename and hash. Publication-integrity validates both snapshot bytes, not historical corpus availability. diff --git a/evidence/indexer-255-covered-cue-20261030/patch/README.md b/evidence/indexer-255-covered-cue-20261030/patch/README.md new file mode 100644 index 000000000..f2c2eb207 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/patch/README.md @@ -0,0 +1 @@ +No candidate runtime patch: both arms used unchanged production6e6ee245; only the frozen relevance cue input differs. diff --git a/evidence/indexer-255-covered-cue-20261030/plan/accepted-receipt.json b/evidence/indexer-255-covered-cue-20261030/plan/accepted-receipt.json new file mode 100644 index 000000000..873eee5d3 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/accepted-receipt.json @@ -0,0 +1,8 @@ +{ + "status": "ACCEPTED_BEFORE_CAPSULES_AND_CASES", + "at_utc": "2026-10-30T03:43:39.735991+00:00", + "protocol_sha256": "4512e1ea3d56235b207e4926c0ed84e8eb3022a461830c919ad29810ee27b026", + "baseline": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "all_six_source_supported_required": true, + "previous_results_unchanged": true +} diff --git a/evidence/indexer-255-covered-cue-20261030/plan/author-capsules.json b/evidence/indexer-255-covered-cue-20261030/plan/author-capsules.json new file mode 100644 index 000000000..2ce338e51 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/author-capsules.json @@ -0,0 +1,44 @@ +{ + "status": "FROZEN_SOURCE_CONDITIONED_CAPSULES_BEFORE_TASK_AUTHORING", + "baseline": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "provenance": "Curator selected covered concrete decisions; labels/titles/catalog IDs omitted. Source passages expose technical terminology/answers, so this is not independent natural-user sampling.", + "instruction": "Use only these capsules and accepted protocol. Author six fresh task/raw-cue pairs, one per capsule, avoid copy-source answer wording in question; two concise controls, at least one exact identity. Do not inspect source map or catalog.", + "capsules": [ + { + "capsule_id": "capsule-01", + "excerpt": "**F**: `memory.high` triggers reclaim/throttling without invoking OOM;\n`memory.max` can invoke OOM; `memory.low` supplies best-effort protection.\n`memory.pressure` exposes PSI; `memory.events.local` excludes descendant\nevents.\n\n**P** triage for host-memory stalls (latency regression):\n1. Align `memory.pressure` (PSI), `memory.events` vs `memory.events.local`\n deltas (high / max / OOM), and the footprint mix of the group.\n2. Reclaim/throttle symptoms \u2192 draft a `memory.high` review.\n3. OOM risk \u2192 explicit `memory.max` review with OOM-risk assessment\n (raising `max` trades crashes for capacity).\n4. Protection needs \u2192 `memory.low` is best-effort only; do not rely on it\n as a guarantee.\n5. Track application outcomes and sibling pressure; additional headroom\n consumes host capacity.", + "boundary": "Proposed diagnosis only: distinguish host memory pressure, limits and best-effort protection; no settings changes or claimed measured benefit.", + "excerpt_sha256": "6b9c52eeea14351003bdba75b5f1dd99b419ef923fc48c27d472cd9f4a5449ce" + }, + { + "capsule_id": "capsule-02", + "excerpt": "Changing the card group is a contract change. It invalidates any previous baseline measurement taken on a different group. After a card-group change:\n- Re-measure the baseline on the new group in the same session window.\n- Do not compare new-group results to old-group baselines (cross-group absolute values are not comparable, for the same reason cross-window measurements are not comparable on a shared machine).", + "boundary": "Measurement-contract question only. Do not initiate device runs, recovery or driver changes.", + "excerpt_sha256": "9f14a05d00c4caf690e145f5cf6448ee9fbe52d4b6f82d6b7b0e7ea56c2d2e46" + }, + { + "capsule_id": "capsule-03", + "excerpt": "| Mode | Semantics | Implementation shape |\n|---|---|---|\n| `distribute` | spread execution evenly over all nodes | round-robin worker threads across nodes (`thread_n % n_nodes`) |\n| `isolate` | only spawn threads on the node execution started on | pin to the current node \u2014 NUMA-local, lowest cross-node traffic |\n| `numactl` | use the CPU map provided by numactl | re-apply the externally captured affinity cpuset |", + "boundary": "llama.cpp Linux NUMA mode semantics only, not a universal fastest-mode claim. No placement mutation.", + "excerpt_sha256": "eb035c5bca1cc4132da484433140331b1d8da9803d12cdfad3e9ffd525720940" + }, + { + "capsule_id": "capsule-04", + "excerpt": "3. **Check replay correctness first.** Require caller-managed stable replay buffers. The wrapper neither owns persistent input buffers nor copies new inputs into them. Propose `VLLM_LOGGING_LEVEL=DEBUG` before wrapper creation for a diagnostic reproduction: its address assertion covers only top-level positional tensor arguments, not keyword/nested inputs or semantic correctness. Escalate an address mismatch to the buffer-owning caller; never suppress it to obtain a timing. [F3]\n4. **Preserve lifetime and ordering.** Review offloader synchronization/join, graph-pool ownership, output/workspace weak references, and the FULL replay/update branch before proposing code changes. Do not remove synchronization, reorder update/replay, or generalize weak-output handling as a performance shortcut. Follow the branch-specific invariants in source-notes. [F4\u2013F5]", + "boundary": "Pinned vLLM-Ascend wrapper source-derived guidance; installed-stack compatibility and runtime efficacy unverified. No execution.", + "excerpt_sha256": "e58c8780c7863b5e2e980c0c82aaaef097740f4be4b2551e8490f50970aa9c63" + }, + { + "capsule_id": "capsule-05", + "excerpt": "3. **Choose a path-specific proposal.** F4: the host allocator is selected once. Default-path round/cache thresholds are explicitly inert with `pin_memory_expandable_segments=True`. P: in fresh-process trials only, consider `pinned_max_round_threshold_mb` for large rounding waste, or `pinned_max_cached_size_mb` for infrequently reused large buffers. Potential benefit is lower retained RAM; costs are less reuse and more allocation/free work. Require the pinned docs' TorchNPU 26.2.0/PyTorch 2.13.0 gate and verify the actual stack. Do not present the latter as a total-cache cap or permission to free in-flight buffers.\n\n4. **Match event-processing and expandable options to symptoms.** P: consider `pinned_use_background_threads` only when foreground event processing is material and its supported default-path implementation is verified; costs include CPU/thread activity. Consider host expandable segments for variable-size pressure only after its separate gates pass; the source uses 20 MiB physical segments, adding granularity/mapping costs. Do not transplant device `expandable_segments`, device split limits or 1 GiB page logic into host settings.", + "boundary": "Source-derived TorchNPU host-RAM allocator policy; gates and effective path must be checked. Not device HBM, total-cache quota or free-inflight authorization.", + "excerpt_sha256": "2dac103907442504227e2b2e3984df18845c4cb7d51aa05d3b2944608875851d" + }, + { + "capsule_id": "capsule-06", + "excerpt": "- Parallelize with bounded workers (`--max-workers 16` style), select only\n needed shards (`--include '{subset}/**'`).\n- `.incomplete` marker files mean in-flight shards, not corruption; wait or\n resume, do not re-download blindly.\n- Verify completeness structurally: shard numbering has no gaps **and** the\n set matches `*.safetensors.index.json` `weight_map` keys. On mismatch,\n cross-check weight_map and fetch only the missing shards.", + "boundary": "Structural download completeness only, not tensor correctness or model efficacy. No actual downloads requested.", + "excerpt_sha256": "10341da8e139eeb169ed90e0656cb655072f8d61ac7c19a38841fb80589280d0" + } + ] +} diff --git a/evidence/indexer-255-covered-cue-20261030/plan/curator-source-labels.json b/evidence/indexer-255-covered-cue-20261030/plan/curator-source-labels.json new file mode 100644 index 000000000..c108d7ff6 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/curator-source-labels.json @@ -0,0 +1,80 @@ +{ + "baseline": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "curated_at_utc": "2026-10-30T03:45:22.798091+00:00", + "selection_basis": "Six different covered engineering decisions, no ranking calls/computation/output inspection. Not selected by measured wins.", + "labels": [ + { + "capsule_id": "capsule-01", + "positive_ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "source_path": "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md", + "source_sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5", + "start_line": 73, + "end_line": 87, + "exact_excerpt": "**F**: `memory.high` triggers reclaim/throttling without invoking OOM;\n`memory.max` can invoke OOM; `memory.low` supplies best-effort protection.\n`memory.pressure` exposes PSI; `memory.events.local` excludes descendant\nevents.\n\n**P** triage for host-memory stalls (latency regression):\n1. Align `memory.pressure` (PSI), `memory.events` vs `memory.events.local`\n deltas (high / max / OOM), and the footprint mix of the group.\n2. Reclaim/throttle symptoms \u2192 draft a `memory.high` review.\n3. OOM risk \u2192 explicit `memory.max` review with OOM-risk assessment\n (raising `max` trades crashes for capacity).\n4. Protection needs \u2192 `memory.low` is best-effort only; do not rely on it\n as a guarantee.\n5. Track application outcomes and sibling pressure; additional headroom\n consumes host capacity.", + "boundary": "Proposed diagnosis only: distinguish host memory pressure, limits and best-effort protection; no settings changes or claimed measured benefit." + }, + { + "capsule_id": "capsule-02", + "positive_ids": [ + "inferrsi.ascend.skills.ascend_device_health" + ], + "source_path": "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md", + "source_sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a", + "start_line": 109, + "end_line": 111, + "exact_excerpt": "Changing the card group is a contract change. It invalidates any previous baseline measurement taken on a different group. After a card-group change:\n- Re-measure the baseline on the new group in the same session window.\n- Do not compare new-group results to old-group baselines (cross-group absolute values are not comparable, for the same reason cross-window measurements are not comparable on a shared machine).", + "boundary": "Measurement-contract question only. Do not initiate device runs, recovery or driver changes." + }, + { + "capsule_id": "capsule-03", + "positive_ids": [ + "inferrsi.common.skills.os_process_placement" + ], + "source_path": "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md", + "source_sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053", + "start_line": 80, + "end_line": 84, + "exact_excerpt": "| Mode | Semantics | Implementation shape |\n|---|---|---|\n| `distribute` | spread execution evenly over all nodes | round-robin worker threads across nodes (`thread_n % n_nodes`) |\n| `isolate` | only spawn threads on the node execution started on | pin to the current node \u2014 NUMA-local, lowest cross-node traffic |\n| `numactl` | use the CPU map provided by numactl | re-apply the externally captured affinity cpuset |", + "boundary": "llama.cpp Linux NUMA mode semantics only, not a universal fastest-mode claim. No placement mutation." + }, + { + "capsule_id": "capsule-04", + "positive_ids": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay" + ], + "source_path": "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md", + "source_sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2", + "start_line": 32, + "end_line": 33, + "exact_excerpt": "3. **Check replay correctness first.** Require caller-managed stable replay buffers. The wrapper neither owns persistent input buffers nor copies new inputs into them. Propose `VLLM_LOGGING_LEVEL=DEBUG` before wrapper creation for a diagnostic reproduction: its address assertion covers only top-level positional tensor arguments, not keyword/nested inputs or semantic correctness. Escalate an address mismatch to the buffer-owning caller; never suppress it to obtain a timing. [F3]\n4. **Preserve lifetime and ordering.** Review offloader synchronization/join, graph-pool ownership, output/workspace weak references, and the FULL replay/update branch before proposing code changes. Do not remove synchronization, reorder update/replay, or generalize weak-output handling as a performance shortcut. Follow the branch-specific invariants in source-notes. [F4\u2013F5]", + "boundary": "Pinned vLLM-Ascend wrapper source-derived guidance; installed-stack compatibility and runtime efficacy unverified. No execution." + }, + { + "capsule_id": "capsule-05", + "positive_ids": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle" + ], + "source_path": "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md", + "source_sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d", + "start_line": 28, + "end_line": 30, + "exact_excerpt": "3. **Choose a path-specific proposal.** F4: the host allocator is selected once. Default-path round/cache thresholds are explicitly inert with `pin_memory_expandable_segments=True`. P: in fresh-process trials only, consider `pinned_max_round_threshold_mb` for large rounding waste, or `pinned_max_cached_size_mb` for infrequently reused large buffers. Potential benefit is lower retained RAM; costs are less reuse and more allocation/free work. Require the pinned docs' TorchNPU 26.2.0/PyTorch 2.13.0 gate and verify the actual stack. Do not present the latter as a total-cache cap or permission to free in-flight buffers.\n\n4. **Match event-processing and expandable options to symptoms.** P: consider `pinned_use_background_threads` only when foreground event processing is material and its supported default-path implementation is verified; costs include CPU/thread activity. Consider host expandable segments for variable-size pressure only after its separate gates pass; the source uses 20 MiB physical segments, adding granularity/mapping costs. Do not transplant device `expandable_segments`, device split limits or 1 GiB page logic into host settings.", + "boundary": "Source-derived TorchNPU host-RAM allocator policy; gates and effective path must be checked. Not device HBM, total-cache quota or free-inflight authorization." + }, + { + "capsule_id": "capsule-06", + "positive_ids": [ + "inferrsi.common.skills.weight_loading_optimization" + ], + "source_path": "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md", + "source_sha256": "a1bf0e2ec066d5f30225c6cc5fb603f830f48eb5e2aee3fc3d084ed9d8aebe2f", + "start_line": 79, + "end_line": 85, + "exact_excerpt": "- Parallelize with bounded workers (`--max-workers 16` style), select only\n needed shards (`--include '{subset}/**'`).\n- `.incomplete` marker files mean in-flight shards, not corruption; wait or\n resume, do not re-download blindly.\n- Verify completeness structurally: shard numbering has no gaps **and** the\n set matches `*.safetensors.index.json` `weight_map` keys. On mismatch,\n cross-check weight_map and fetch only the missing shards.", + "boundary": "Structural download completeness only, not tensor correctness or model efficacy. No actual downloads requested." + } + ], + "overlap": "Coordinator recalled historical v1 DEV CPU-quota and weight-loading mmap cases. Before author delivery and any scoring, capsule01 changed from CPU quota to host-memory pressure/high-max-low semantics, capsule03 changed from affinity-measurement discipline to Linux NUMA mode semantics to distinguish from card-group contract and omit title crosslink. Same cgroup and weight-loading owner families remain familiar; weight-download completeness differs from mmap. This is source-conditioned DEV, not independent heldout." +} diff --git a/evidence/indexer-255-covered-cue-20261030/plan/evaluation-inputs.json b/evidence/indexer-255-covered-cue-20261030/plan/evaluation-inputs.json new file mode 100644 index 000000000..83f0e87c0 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/evaluation-inputs.json @@ -0,0 +1,125 @@ +{ + "baseline": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "cases": [ + { + "id": "covered255-01", + "raw_task": "Could you help with this investigation? Help me distinguish throttling, an out-of-memory limit, and memory protection when investigating a latency regression. The affected inference service runs in a Linux memory cgroup, and I can collect host-memory pressure and event counters for that group and its children. Give a diagnostic plan that separates local events from child activity and considers pressure on sibling workloads; explain the limits of any protection guarantee, but do not change settings or claim a measured improvement.", + "cue": "Help me distinguish throttling, an out-of-memory limit, and memory protection when investigating a latency regression.; The affected inference service runs in a Linux memory cgroup, and I can collect host-memory pressure and event counters for that group and its children.; Give a diagnostic plan that separates local events from child activity and considers pressure on sibling workloads; explain the limits of any protection guarantee, but do not change settings or claim a measured improvement.", + "identity_control": false, + "query": "", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "evidence_path": "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md", + "source_sha256": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5", + "source_lines": [ + 73, + 87 + ], + "source_limits": "Proposed diagnosis only: distinguish host memory pressure, limits and best-effort protection; no settings changes or claimed measured benefit." + }, + { + "id": "covered255-02", + "raw_task": "Decide whether our before-and-after performance comparison is valid and specify what evidence is needed; an accelerator experiment whose assigned card group changed between baseline and candidate measurements on a shared machine; provide a measurement plan only; do not run devices or change drivers", + "cue": "Decide whether our before-and-after performance comparison is valid and specify what evidence is needed; an accelerator experiment whose assigned card group changed between baseline and candidate measurements on a shared machine; provide a measurement plan only; do not run devices or change drivers", + "identity_control": true, + "query": "", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.ascend.skills.ascend_device_health" + ], + "evidence_path": "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md", + "source_sha256": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a", + "source_lines": [ + 109, + 111 + ], + "source_limits": "Measurement-contract question only. Do not initiate device runs, recovery or driver changes." + }, + { + "id": "covered255-03", + "raw_task": "Explain which placement mode respects an externally supplied CPU affinity map and how the alternatives differ; llama.cpp on Linux with NUMA modes distribute, isolate, and numactl; describe thread placement only; do not change affinity or claim one mode is universally fastest", + "cue": "Explain which placement mode respects an externally supplied CPU affinity map and how the alternatives differ; llama.cpp on Linux with NUMA modes distribute, isolate, and numactl; describe thread placement only; do not change affinity or claim one mode is universally fastest", + "identity_control": true, + "query": "", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.common.skills.os_process_placement" + ], + "evidence_path": "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md", + "source_sha256": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053", + "source_lines": [ + 80, + 84 + ], + "source_limits": "llama.cpp Linux NUMA mode semantics only, not a universal fastest-mode claim. No placement mutation." + }, + { + "id": "covered255-04", + "raw_task": "Please help me review this. Prepare a correctness review for replaying a captured graph with a new input and identify what a diagnostic reproduction could and could not establish. The component is a vLLM-Ascend graph wrapper, and the call can contain positional tensors as well as nested or keyword inputs. Check who is responsible for input storage and updates, include lifetime and ordering concerns, and keep this source-based rather than claiming installed-stack compatibility; do not run the reproduction or trade correctness checks for a better timing.", + "cue": "Prepare a correctness review for replaying a captured graph with a new input and identify what a diagnostic reproduction could and could not establish.; The component is a vLLM-Ascend graph wrapper, and the call can contain positional tensors as well as nested or keyword inputs.; Check who is responsible for input storage and updates, include lifetime and ordering concerns, and keep this source-based rather than claiming installed-stack compatibility; do not run the reproduction or trade correctness checks for a better timing.", + "identity_control": false, + "query": "", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay" + ], + "evidence_path": "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md", + "source_sha256": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2", + "source_lines": [ + 32, + 33 + ], + "source_limits": "Pinned vLLM-Ascend wrapper source-derived guidance; installed-stack compatibility and runtime efficacy unverified. No execution." + }, + { + "id": "covered255-05", + "raw_task": "Could you advise on this memory issue? Propose how to investigate retained host RAM from large, rarely reused pinned buffers and decide which allocator options are relevant. The workload uses TorchNPU, and we have not yet verified the installed versions or whether pin_memory_expandable_segments=True is active. First establish the applicable version and allocator-path gates, describe the trade-offs of any proposed settings in fresh-process trials, and distinguish host allocation from device memory; do not run trials, promise a total-cache quota, or free buffers that are still in flight.", + "cue": "Propose how to investigate retained host RAM from large, rarely reused pinned buffers and decide which allocator options are relevant.; The workload uses TorchNPU, and we have not yet verified the installed versions or whether pin_memory_expandable_segments=True is active.; First establish the applicable version and allocator-path gates, describe the trade-offs of any proposed settings in fresh-process trials, and distinguish host allocation from device memory; do not run trials, promise a total-cache quota, or free buffers that are still in flight.", + "identity_control": false, + "query": "", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle" + ], + "evidence_path": "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md", + "source_sha256": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d", + "source_lines": [ + 28, + 30 + ], + "source_limits": "Source-derived TorchNPU host-RAM allocator policy; gates and effective path must be checked. Not device HBM, total-cache quota or free-inflight authorization." + }, + { + "id": "covered255-06", + "raw_task": "Please help me check this download. Explain how to tell whether a sharded model download is complete and how to recover only what is missing. The download directory contains numbered safetensors shards, an index JSON, and some .incomplete marker files. Base the assessment on shard structure and the index rather than assuming the markers prove corruption; describe a bounded-concurrency recovery plan without downloading anything, and do not treat structural completeness as proof of tensor correctness or model quality.", + "cue": "Explain how to tell whether a sharded model download is complete and how to recover only what is missing.; The download directory contains numbered safetensors shards, an index JSON, and some .incomplete marker files.; Base the assessment on shard structure and the index rather than assuming the markers prove corruption; describe a bounded-concurrency recovery plan without downloading anything, and do not treat structural completeness as proof of tensor correctness or model quality.", + "identity_control": false, + "query": "", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.common.skills.weight_loading_optimization" + ], + "evidence_path": "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md", + "source_sha256": "a1bf0e2ec066d5f30225c6cc5fb603f830f48eb5e2aee3fc3d084ed9d8aebe2f", + "source_lines": [ + 79, + 85 + ], + "source_limits": "Structural download completeness only, not tensor correctness or model efficacy. No actual downloads requested." + } + ] +} diff --git a/evidence/indexer-255-covered-cue-20261030/plan/input-freeze.json b/evidence/indexer-255-covered-cue-20261030/plan/input-freeze.json new file mode 100644 index 000000000..e6348b737 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/input-freeze.json @@ -0,0 +1,15 @@ +{ + "frozen_at_utc": "2026-10-30T03:47:12.322822+00:00", + "status": "ALL_INPUTS_FROZEN_BEFORE_RETRIEVAL; RUNNER_FREEZE_STILL_REQUIRED", + "files": { + "accepted-receipt.json": "3b4bd0536b1e99247314a8788238de9b12e0ab0ccaef8e52941394f2b08bc384", + "author-capsules.json": "5fa54fb0ec903c9839dc1e1d26e14cf8469c7ca7cd79843bd5147e7ca901b822", + "curator-source-labels.json": "b2e1dfde979206803058afaa9b7e4d4e75633bba7259a5f130eb35f5ac36eb71", + "evaluation-inputs.json": "24e822eaadcd4598435dac0e7cf77cc1dd73d8a6c50bc0f7c5982253b4a94fcc", + "input-review.json": "1e5eb6392e3e879524ce223c63673c8f495bbd8e97d0de35f165508eb557306c", + "protocol-draft.md": "4512e1ea3d56235b207e4926c0ed84e8eb3022a461830c919ad29810ee27b026", + "source-conditioned-author.json": "e8a52fb88791aa5a1cbe079d9629540dcd104894185955370bd3d65cb842b568", + "source-conditioned-author.json.sha256": "97674d5214fbd6c8dd672967677a43342abc66e6958a294434a73153c3e66e4a", + "source-manifest.json": "67a40dc08dcab6f8cafd008a3005e371b1a77bdc31e08f75b8c14335dfa60b56" + } +} diff --git a/evidence/indexer-255-covered-cue-20261030/plan/input-review.json b/evidence/indexer-255-covered-cue-20261030/plan/input-review.json new file mode 100644 index 000000000..36d387def --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/input-review.json @@ -0,0 +1,16 @@ +{ + "status": "PASS_SIX_INTENT_AND_SOURCE_SUPPORTED_BEFORE_RETRIEVAL", + "reviewed_at_utc": "2026-10-30T03:47:12.322312+00:00", + "baseline": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "author_input_sha256": "e8a52fb88791aa5a1cbe079d9629540dcd104894185955370bd3d65cb842b568", + "supported_count": 6, + "identity_controls": 2, + "no_input_or_label_edits": true, + "limits": [ + "All6 are source-conditioned synthetic cases, not unbiased user traffic.", + "Fourchanged pairs remove politeness only; not substantive narrative compression.", + "Card-group case uses generally transferable fixed-hardware comparison principle from Ascend-oriented source; not generic device/runtime certification.", + "Same owner-family familiarity for historical cgroupCPUquota and weightloadingmmap; mechanisms differ and no outcome-based selection used.", + "Partial positives not complete ground truth. Oldsealedheldout not accessed/executed." + ] +} diff --git a/evidence/indexer-255-covered-cue-20261030/plan/protocol-draft.md b/evidence/indexer-255-covered-cue-20261030/plan/protocol-draft.md new file mode 100644 index 000000000..a99324b5e --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/protocol-draft.md @@ -0,0 +1,35 @@ +# #255 source-conditioned cue feasibility — new prospective study + +Status: NEW PROTOCOL FOR COORDINATOR CONFIRMATION BEFORE CASE PREPARATION/EVALUATION. Owner continuation03:41 UTC2026-10-30. This does not replace or repair the preceding blind-task study's preparation-INCONCLUSIVE result (three supported pairs below its four-case minimum); retain that result and its exact inputs separately. + +## Identity and narrow question +Freshly fetched main `6e6ee245d9e92ee9365976f2e4e262304999f147` includes `knowledge(hardware): audit sources and deepen memory, graph and collective guidance (#262)`, parent `b3549ce1`. Pin this new baseline explicitly. Both arms must use this same corpus and unchanged production weighted-lexical-v1 runtime. The earlier study remains on its historicalb3549ce1; do not reattribute it to this new corpus. + +Question: on six deliberately source-covered synthetic tasks, does manual task-preserving operation/object/constraint extraction improve ranking relative to the full task narrative? This tests a conditional input-contract feasibility hypothesis, NOT organic user demand, coverage of the whole repository, unbiased retrieval accuracy or automatic rewriting. No new code/scorer, paid/model service, hardware experiment, broad service or default change. + +## Role-separated preparation +1. **Source curator** reads pinned source and selects exactly six distinct, covered task intents. For each, freeze partial positive skill IDs, exact source passages/line ranges and SHA256, accepted boundary and limits. Choose for distinct concrete engineering decisions, not keyword density or expected ranking. No retrieval calls, scorer computation, output inspection or choosing based on previous candidate results. Avoid old observed DEV and old sealed-heldout intents when possible; a coordinator may screen overlap without exposing sealed text to implementer or task author. Disclose unavoidable overlap as reduced independence rather than silently substituting after outcomes. +2. **Independent task author** receives only the six curated source passages and boundary limits stripped of skill titles, catalog IDs, file paths and metadata search tags. The author may see necessary technical API/device names in evidence; these cannot honestly be called blinded. No catalog access, retrieval ranks or previous task/outcome material. Write one plausible task per source capsule, then apply the exact extraction rule below. Do not copy source wording merely to match the index, leak the answer in the task, or invent unsupported capability. This is SOURCE-CONDITIONED synthetic authoring, not independent real-user sampling or blind-to-source labelling. +3. **Intent/label reviewer** checks task-to-source support and raw/cue intent equivalence before retrieval; source-positive IDs are those prospectively selected, not retrofitted from outputs. If a task is unsupported, changes the stated scope or fails intent preservation, stop preparation as INCONCLUSIVE; do not replace it to ensure a passing score. A factual pre-evaluation correction must be explicit and requires coordinator decision and refreeze before any run; no post-output edits. +4. Freeze all six complete raw/cue pairs, span maps, source labels/quotes/hashes, overlap/access attestations, runtime/source manifest and evaluation runner before any retrieval. Independent evaluator performs one paired CPU run. Old heldout content remains unused and undisclosed; no new mode or candidate is implemented. + +## Unchanged manual extraction rule +Select verbatim OPERATION, OBJECT and ALL CONSTRAINTS spans from the authored raw task. Join with `; ` in that order. Preserve uncertainty, environment/API identity, thresholds, negation, safety and requested deliverable. Remove only greetings, politeness, redundant framing or incidental chronology not used by the task; record every removed span and its permitted reason. No added keywords, synonyms, inflection, translation, schema-keywords or answer terms. Do not concatenate separately selected words to invent a new phrase. This is deterministic serialization of manually selected spans, not an automatic semantic extractor. + +Do not artificially pad tasks or bait substring matches. Include at least two naturally concise controls among six, with at least one raw/cue byte-identical. If all meaningful wording must be kept, keep it. Record task-author effort, character/whitespace-token counts and whether the only change was polite framing; do not claim broad narrative normalization if that is all this set tests. + +## Prospective evaluation and fixed gates +All six tasks must be source-supported and intent-valid before running. If any remains unsupported, preparation is INCONCLUSIVE without a retrieval attempt. This deliberately conditioned all-six coverage requirement is separate from the previous blind study's four-of-six threshold; it does not alter that study. + +Both arms: `query=''`, no filters, `limit=5`, `context_bytes=262144`, production `relevance` API; only the cue string differs. Runtime and source are identical. One capture per arm/task, no rewording or rerun after outcomes. Freeze runner before use; preserve all raw outputs, misses, resource omissions and errors. + +- No raw-arm labelled Hit@5 may be lost; Hit@5 nondecreasing; MRR@5 STRICTLY greater than raw baseline across all six. Report exact ranks and arithmetic without rounding into a pass. Labels are partial positives, not complete precision/recall truth. +- Byte-identical controls produce identical serialized output. Near-identity controls are reported descriptively rather than falsely requiring parity. +- Existing six default/unchanged-interface contract captures retain exact parity on the same pinned runtime, and no production-source diff is introduced. +- Full indexed_skill required-resource set, exact emitted bytes/hash/revision, budget accounting and explicit omissions pass in each arm; complete relevant delivery count must not decrease. Minimum suffixes are anchors, not exhaustive required sets. +- Any complete gate failure = FAIL; source/identity/execution incompleteness = INCONCLUSIVE. Stop, retain evidence and do not try another cue rule or swap cases until a separate prospective decision. + +## Interpretation +PASS would support only this manually prepared source-conditioned set. It cannot establish that arbitrary users phrase questions this way, the repository covers their tasks, human preparation is cheap, or an automatic normalizer works. Evidence capsules disclose answer/source familiarity as a design limitation; role separation reduces direct title/ID leakage but does not remove source-conditioning bias. Do not pool the six cases with earlier DEV or reuse their scores as independent replications. No timing or downstream task-success claim, no promotion/default change or automatic #255 closure. + +Previous performance/v2/v3 FAIL and blind-input preparation-INCONCLUSIVE remain visible. Preserve the old sealed six heldout cases untouched. diff --git a/evidence/indexer-255-covered-cue-20261030/plan/publication-integrity.json b/evidence/indexer-255-covered-cue-20261030/plan/publication-integrity.json new file mode 100644 index 000000000..e9a0aaffd --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/publication-integrity.json @@ -0,0 +1,36 @@ +{ + "README.md": "1b5e52d48db33641f7b20c5a224c4a807c20a2e34d6d4682d90845a60d0cf102", + "patch/README.md": "27335a4b5468dfb065d82c16c4431404a3208028c87cb4cf0c7ba0bc1329fa12", + "plan/accepted-receipt.json": "3b4bd0536b1e99247314a8788238de9b12e0ab0ccaef8e52941394f2b08bc384", + "plan/author-capsules.json": "5fa54fb0ec903c9839dc1e1d26e14cf8469c7ca7cd79843bd5147e7ca901b822", + "plan/curator-source-labels.json": "b2e1dfde979206803058afaa9b7e4d4e75633bba7259a5f130eb35f5ac36eb71", + "plan/evaluation-inputs.json": "24e822eaadcd4598435dac0e7cf77cc1dd73d8a6c50bc0f7c5982253b4a94fcc", + "plan/input-freeze.json": "305246e2c49ea47bea9dbe9dc412e5a8920acec735f01f445514938b8cf7b9c3", + "plan/input-review.json": "1e5eb6392e3e879524ce223c63673c8f495bbd8e97d0de35f165508eb557306c", + "plan/protocol-draft.md": "4512e1ea3d56235b207e4926c0ed84e8eb3022a461830c919ad29810ee27b026", + "plan/runner-freeze.json": "9363d36e0a21703916c7d7d9f2558d54d675e89299d05ca4ef705429ca3b2d52", + "plan/source-conditioned-author.json": "e8a52fb88791aa5a1cbe079d9629540dcd104894185955370bd3d65cb842b568", + "plan/source-conditioned-author.json.sha256": "97674d5214fbd6c8dd672967677a43342abc66e6958a294434a73153c3e66e4a", + "plan/source-pins.json": "67a40dc08dcab6f8cafd008a3005e371b1a77bdc31e08f75b8c14335dfa60b56", + "results/consolidated-results.json": "12792f311610fa291cadd6552cb3dc9929fe5e6bb017c53854c27db8a6d95f7a", + "results/consolidated-results.md": "6f5de8b61ff105db004ac68564ede561a866b3244c58789fcda736ed0efd7efd", + "results/covered255-01-cue.json.gz": "7c7fdeba31a7e2fe8e2236e04d15e61c377997de95c9698a9f47847cfb49f45d", + "results/covered255-01-raw.json.gz": "f3c85e4711ca4c8db6a81f7a29e72f5476a4421ce1d63e108ac66c330cf0f60d", + "results/covered255-02-cue.json.gz": "5e519f408f1dc23cbd92c91e0cd31c5552a7211f93b4ca64827a9d4a27a444e2", + "results/covered255-02-raw.json.gz": "5e519f408f1dc23cbd92c91e0cd31c5552a7211f93b4ca64827a9d4a27a444e2", + "results/covered255-03-cue.json.gz": "cca479a403d81671614a16cbb091f461e527807340e0158b1d30d8854800109f", + "results/covered255-03-raw.json.gz": "cca479a403d81671614a16cbb091f461e527807340e0158b1d30d8854800109f", + "results/covered255-04-cue.json.gz": "d74eb8080922b13ccf5f94cb4018953015b997bdc807148c812b9f44deb190bd", + "results/covered255-04-raw.json.gz": "7708794d729840f591c9bb5cbb91bbd5a79017b862ec93678a526a15ca8a1260", + "results/covered255-05-cue.json.gz": "c0534b72099981769c26f09393c5f120a8e038b4bf3c0f32774eb095e2abb877", + "results/covered255-05-raw.json.gz": "a7a6c946db65253085fcb67838f98a2f801d2099b9673e51a3ee35d9ffadb555", + "results/covered255-06-cue.json.gz": "11de82eb96df6c932cf52e5b4b67b7c82544b9883d686e89dab23a5f846c157e", + "results/covered255-06-raw.json.gz": "225e31bf129e0ad88fc070b79a7a8da07cb073019d3cc5f97062b61b9dbbb4b6", + "results/default-captures.json": "c40964f7296ef7e6d387bdc569b9a49fc93839010e79a66da426d28035b05d1e", + "results/independent-review.json": "3c57455dda4357b357f59e68095388a3e6859fd27c5b0fa7ad64ed1d40786b09", + "results/partial-results.json": "93caca3d01f7cce09d6d7fa78ba429cde16a860e9b6561d4cfd825be6a57b07c", + "results/report.md": "7af98da8eba7319c5d853b3c7932ab54dd434df50edb221c5de15aa98e7e7bd8", + "results/results.json": "5ef55a52de4ae32909881e5be4101f0d5a2eb8904aaa1185df9d798d99acacba", + "results/run.log": "869085edd6a815272527ff651fdd5eec4953ac44ffa167dbcc555ac2fd8cb95a", + "validator/evaluate.py": "566c9e2d3a63eb3d9b7e8f4b1e41ce1c5121e316c0ffa50ffc01fd0aca40c1af" +} diff --git a/evidence/indexer-255-covered-cue-20261030/plan/runner-freeze.json b/evidence/indexer-255-covered-cue-20261030/plan/runner-freeze.json new file mode 100644 index 000000000..766498435 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/runner-freeze.json @@ -0,0 +1,8 @@ +{ + "frozen_at_utc": "2026-10-30T03:48:25.146092+00:00", + "runtime": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "runner_sha256": "566c9e2d3a63eb3d9b7e8f4b1e41ce1c5121e316c0ffa50ffc01fd0aca40c1af", + "inputs_sha256": "24e822eaadcd4598435dac0e7cf77cc1dd73d8a6c50bc0f7c5982253b4a94fcc", + "input_freeze_sha256": "305246e2c49ea47bea9dbe9dc412e5a8920acec735f01f445514938b8cf7b9c3", + "retrieval_started": false +} diff --git a/evidence/indexer-255-covered-cue-20261030/plan/source-conditioned-author.json b/evidence/indexer-255-covered-cue-20261030/plan/source-conditioned-author.json new file mode 100644 index 000000000..e55efe34c --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/source-conditioned-author.json @@ -0,0 +1,409 @@ +{ + "schema": "source-conditioned-manual-cue-authoring-v1", + "status": "FROZEN_SOURCE_CONDITIONED_SYNTHETIC_INPUTS_BEFORE_INTENT_SOURCE_AUDIT", + "frozen_at_utc": "2026-10-30T03:46:41.865109+00:00", + "baseline": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "protocol_sha256": "4512e1ea3d56235b207e4926c0ed84e8eb3022a461830c919ad29810ee27b026", + "capsules_sha256": "5fa54fb0ec903c9839dc1e1d26e14cf8469c7ca7cd79843bd5147e7ca901b822", + "provenance": "Six fresh tasks, each grounded in one prospectively supplied sanitized source capsule. Source familiarity is explicit. Not blind source labelling, observed user traffic, independent real-user sampling, or evidence about the user-query distribution. Earlier blind-study artifacts are unchanged and not pooled or replaced.", + "access_attestation": { + "read_for_this_study": [ + "protocol-draft.md", + "author-capsules.json", + "Coordinator and curator instructions without source IDs or ranking results" + ], + "source_familiarity": "Author saw technical names, facts, boundaries and potential answers in all six source excerpts. Sanitizing titles/IDs does not remove source-conditioning bias.", + "prior_context": "This same agent authored the separate blind-input set. New protocol/coordinator instructions disclose its aggregate preparation-INCONCLUSIVE status. No prior per-case source labels, ranked outputs, old DEV cases or sealed heldout content were read. No prior tasks were reused.", + "not_accessed": [ + "Catalog or repository source", + "curator-source-labels", + "Skill titles or catalog IDs", + "Scorer implementation", + "Retrieval calls or outputs", + "Model services" + ], + "basis": "Self-reported process history, not forensic isolation." + }, + "manual_rule": "Verbatim OPERATION, OBJECT, and all CONSTRAINTS spans joined with semicolon-space. Deterministic serialization after manual span selection, not automatic semantic extraction. No cue synonyms, new terms, answer hints, translation or catalog-informed edits.", + "offset_convention": "Zero-based Unicode code-point offsets, end-exclusive. All raw text is ASCII. Selected and removed spans exactly partition the raw text; original separators are recorded as removed and the serialization delimiter inserted.", + "effort": { + "started_at_utc": "2026-10-30T03:45:30+00:00", + "frozen_at_utc": "2026-10-30T03:46:41.865109+00:00", + "elapsed_seconds": 71.865109, + "work": "One six-task authoring pass followed by programmatic source-capsule hash, span partition, verbatim serialization and identity checks. No retrieval or outcome revisions." + }, + "interpretation_limit": "The four non-control changes remove polite framing only and change separators. This set does not establish broad narrative normalization. Semantic information was not compressed.", + "cases": [ + { + "case_id": "covered255-01", + "capsule_id": "capsule-01", + "source_excerpt_sha256": "6b9c52eeea14351003bdba75b5f1dd99b419ef923fc48c27d472cd9f4a5449ce", + "intent": "Diagnose host-memory latency symptoms", + "raw_task": "Could you help with this investigation? Help me distinguish throttling, an out-of-memory limit, and memory protection when investigating a latency regression. The affected inference service runs in a Linux memory cgroup, and I can collect host-memory pressure and event counters for that group and its children. Give a diagnostic plan that separates local events from child activity and considers pressure on sibling workloads; explain the limits of any protection guarantee, but do not change settings or claim a measured improvement.", + "cue": "Help me distinguish throttling, an out-of-memory limit, and memory protection when investigating a latency regression.; The affected inference service runs in a Linux memory cgroup, and I can collect host-memory pressure and event counters for that group and its children.; Give a diagnostic plan that separates local events from child activity and considers pressure on sibling workloads; explain the limits of any protection guarantee, but do not change settings or claim a measured improvement.", + "identity_control": false, + "spans": [ + { + "slot": "OPERATION", + "start": 40, + "end": 158, + "text": "Help me distinguish throttling, an out-of-memory limit, and memory protection when investigating a latency regression." + }, + { + "slot": "OBJECT", + "start": 159, + "end": 311, + "text": "The affected inference service runs in a Linux memory cgroup, and I can collect host-memory pressure and event counters for that group and its children." + }, + { + "slot": "CONSTRAINTS", + "start": 312, + "end": 535, + "text": "Give a diagnostic plan that separates local events from child activity and considers pressure on sibling workloads; explain the limits of any protection guarantee, but do not change settings or claim a measured improvement." + } + ], + "removed_spans": [ + { + "start": 0, + "end": 40, + "text": "Could you help with this investigation? ", + "reason": "politeness", + "detail": "Polite request opening." + }, + { + "start": 158, + "end": 159, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + }, + { + "start": 311, + "end": 312, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + } + ], + "constraints_retained": [ + "Give a diagnostic plan that separates local events from child activity and considers pressure on sibling workloads; explain the limits of any protection guarantee, but do not change settings or claim a measured improvement." + ], + "counts": { + "raw_characters": 535, + "cue_characters": 497, + "raw_whitespace_tokens": 80, + "cue_whitespace_tokens": 74 + }, + "raw_sha256": "00afcad041bb0d08b55e214038337d1b7328df5c686f160c377578eac9f308f0", + "cue_sha256": "f1e1015d23c46651683b22f691edf082fda0a178ddc453fef21b3692ce439464", + "compression_scope": "Only polite framing removed and original slot whitespace replaced; all substantive text retained." + }, + { + "case_id": "covered255-02", + "capsule_id": "capsule-02", + "source_excerpt_sha256": "9f14a05d00c4caf690e145f5cf6448ee9fbe52d4b6f82d6b7b0e7ea56c2d2e46", + "intent": "Assess a changed measurement contract", + "raw_task": "Decide whether our before-and-after performance comparison is valid and specify what evidence is needed; an accelerator experiment whose assigned card group changed between baseline and candidate measurements on a shared machine; provide a measurement plan only; do not run devices or change drivers", + "cue": "Decide whether our before-and-after performance comparison is valid and specify what evidence is needed; an accelerator experiment whose assigned card group changed between baseline and candidate measurements on a shared machine; provide a measurement plan only; do not run devices or change drivers", + "identity_control": true, + "spans": [ + { + "slot": "OPERATION", + "start": 0, + "end": 103, + "text": "Decide whether our before-and-after performance comparison is valid and specify what evidence is needed" + }, + { + "slot": "OBJECT", + "start": 105, + "end": 228, + "text": "an accelerator experiment whose assigned card group changed between baseline and candidate measurements on a shared machine" + }, + { + "slot": "CONSTRAINTS", + "start": 230, + "end": 299, + "text": "provide a measurement plan only; do not run devices or change drivers" + } + ], + "removed_spans": [ + { + "start": 103, + "end": 105, + "text": "; ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + }, + { + "start": 228, + "end": 230, + "text": "; ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + } + ], + "constraints_retained": [ + "provide a measurement plan only; do not run devices or change drivers" + ], + "counts": { + "raw_characters": 299, + "cue_characters": 299, + "raw_whitespace_tokens": 43, + "cue_whitespace_tokens": 43 + }, + "raw_sha256": "be5b0bde412e5264be433d9b8739ef428f4465f22b17af7ca2bc027974273675", + "cue_sha256": "be5b0bde412e5264be433d9b8739ef428f4465f22b17af7ca2bc027974273675", + "compression_scope": "Byte-identical control." + }, + { + "case_id": "covered255-03", + "capsule_id": "capsule-03", + "source_excerpt_sha256": "eb035c5bca1cc4132da484433140331b1d8da9803d12cdfad3e9ffd525720940", + "intent": "Interpret NUMA placement options", + "raw_task": "Explain which placement mode respects an externally supplied CPU affinity map and how the alternatives differ; llama.cpp on Linux with NUMA modes distribute, isolate, and numactl; describe thread placement only; do not change affinity or claim one mode is universally fastest", + "cue": "Explain which placement mode respects an externally supplied CPU affinity map and how the alternatives differ; llama.cpp on Linux with NUMA modes distribute, isolate, and numactl; describe thread placement only; do not change affinity or claim one mode is universally fastest", + "identity_control": true, + "spans": [ + { + "slot": "OPERATION", + "start": 0, + "end": 109, + "text": "Explain which placement mode respects an externally supplied CPU affinity map and how the alternatives differ" + }, + { + "slot": "OBJECT", + "start": 111, + "end": 178, + "text": "llama.cpp on Linux with NUMA modes distribute, isolate, and numactl" + }, + { + "slot": "CONSTRAINTS", + "start": 180, + "end": 275, + "text": "describe thread placement only; do not change affinity or claim one mode is universally fastest" + } + ], + "removed_spans": [ + { + "start": 109, + "end": 111, + "text": "; ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + }, + { + "start": 178, + "end": 180, + "text": "; ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + } + ], + "constraints_retained": [ + "describe thread placement only; do not change affinity or claim one mode is universally fastest" + ], + "counts": { + "raw_characters": 275, + "cue_characters": 275, + "raw_whitespace_tokens": 41, + "cue_whitespace_tokens": 41 + }, + "raw_sha256": "824eab8170070bd47e061f025c5b21059673316f7dcc7ef03ff8d34a2916c14f", + "cue_sha256": "824eab8170070bd47e061f025c5b21059673316f7dcc7ef03ff8d34a2916c14f", + "compression_scope": "Byte-identical control." + }, + { + "case_id": "covered255-04", + "capsule_id": "capsule-04", + "source_excerpt_sha256": "e58c8780c7863b5e2e980c0c82aaaef097740f4be4b2551e8490f50970aa9c63", + "intent": "Review graph replay correctness before optimization", + "raw_task": "Please help me review this. Prepare a correctness review for replaying a captured graph with a new input and identify what a diagnostic reproduction could and could not establish. The component is a vLLM-Ascend graph wrapper, and the call can contain positional tensors as well as nested or keyword inputs. Check who is responsible for input storage and updates, include lifetime and ordering concerns, and keep this source-based rather than claiming installed-stack compatibility; do not run the reproduction or trade correctness checks for a better timing.", + "cue": "Prepare a correctness review for replaying a captured graph with a new input and identify what a diagnostic reproduction could and could not establish.; The component is a vLLM-Ascend graph wrapper, and the call can contain positional tensors as well as nested or keyword inputs.; Check who is responsible for input storage and updates, include lifetime and ordering concerns, and keep this source-based rather than claiming installed-stack compatibility; do not run the reproduction or trade correctness checks for a better timing.", + "identity_control": false, + "spans": [ + { + "slot": "OPERATION", + "start": 28, + "end": 179, + "text": "Prepare a correctness review for replaying a captured graph with a new input and identify what a diagnostic reproduction could and could not establish." + }, + { + "slot": "OBJECT", + "start": 180, + "end": 306, + "text": "The component is a vLLM-Ascend graph wrapper, and the call can contain positional tensors as well as nested or keyword inputs." + }, + { + "slot": "CONSTRAINTS", + "start": 307, + "end": 558, + "text": "Check who is responsible for input storage and updates, include lifetime and ordering concerns, and keep this source-based rather than claiming installed-stack compatibility; do not run the reproduction or trade correctness checks for a better timing." + } + ], + "removed_spans": [ + { + "start": 0, + "end": 28, + "text": "Please help me review this. ", + "reason": "politeness", + "detail": "Polite request opening." + }, + { + "start": 179, + "end": 180, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + }, + { + "start": 306, + "end": 307, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + } + ], + "constraints_retained": [ + "Check who is responsible for input storage and updates, include lifetime and ordering concerns, and keep this source-based rather than claiming installed-stack compatibility; do not run the reproduction or trade correctness checks for a better timing." + ], + "counts": { + "raw_characters": 558, + "cue_characters": 532, + "raw_whitespace_tokens": 86, + "cue_whitespace_tokens": 81 + }, + "raw_sha256": "ba303f534fa4b61eddf9f9cff550ff131638b17d8d86ea5aff9da4de252d0824", + "cue_sha256": "057361325d0d9dcf3ec851c66c4545cdeb2fdff5abd6912307b41b1b27559846", + "compression_scope": "Only polite framing removed and original slot whitespace replaced; all substantive text retained." + }, + { + "case_id": "covered255-05", + "capsule_id": "capsule-05", + "source_excerpt_sha256": "2dac103907442504227e2b2e3984df18845c4cb7d51aa05d3b2944608875851d", + "intent": "Propose gated host allocator investigation", + "raw_task": "Could you advise on this memory issue? Propose how to investigate retained host RAM from large, rarely reused pinned buffers and decide which allocator options are relevant. The workload uses TorchNPU, and we have not yet verified the installed versions or whether pin_memory_expandable_segments=True is active. First establish the applicable version and allocator-path gates, describe the trade-offs of any proposed settings in fresh-process trials, and distinguish host allocation from device memory; do not run trials, promise a total-cache quota, or free buffers that are still in flight.", + "cue": "Propose how to investigate retained host RAM from large, rarely reused pinned buffers and decide which allocator options are relevant.; The workload uses TorchNPU, and we have not yet verified the installed versions or whether pin_memory_expandable_segments=True is active.; First establish the applicable version and allocator-path gates, describe the trade-offs of any proposed settings in fresh-process trials, and distinguish host allocation from device memory; do not run trials, promise a total-cache quota, or free buffers that are still in flight.", + "identity_control": false, + "spans": [ + { + "slot": "OPERATION", + "start": 39, + "end": 173, + "text": "Propose how to investigate retained host RAM from large, rarely reused pinned buffers and decide which allocator options are relevant." + }, + { + "slot": "OBJECT", + "start": 174, + "end": 311, + "text": "The workload uses TorchNPU, and we have not yet verified the installed versions or whether pin_memory_expandable_segments=True is active." + }, + { + "slot": "CONSTRAINTS", + "start": 312, + "end": 592, + "text": "First establish the applicable version and allocator-path gates, describe the trade-offs of any proposed settings in fresh-process trials, and distinguish host allocation from device memory; do not run trials, promise a total-cache quota, or free buffers that are still in flight." + } + ], + "removed_spans": [ + { + "start": 0, + "end": 39, + "text": "Could you advise on this memory issue? ", + "reason": "politeness", + "detail": "Polite request opening." + }, + { + "start": 173, + "end": 174, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + }, + { + "start": 311, + "end": 312, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + } + ], + "constraints_retained": [ + "First establish the applicable version and allocator-path gates, describe the trade-offs of any proposed settings in fresh-process trials, and distinguish host allocation from device memory; do not run trials, promise a total-cache quota, or free buffers that are still in flight." + ], + "counts": { + "raw_characters": 592, + "cue_characters": 555, + "raw_whitespace_tokens": 86, + "cue_whitespace_tokens": 79 + }, + "raw_sha256": "05c6f3800ea5435bc40620175ac1e8861984e071643f0c723ae546d22cfd25d8", + "cue_sha256": "63f839a4d76b92359e7e082c1373885800ee7da2e94731a2d40ca49f50d166d3", + "compression_scope": "Only polite framing removed and original slot whitespace replaced; all substantive text retained." + }, + { + "case_id": "covered255-06", + "capsule_id": "capsule-06", + "source_excerpt_sha256": "10341da8e139eeb169ed90e0656cb655072f8d61ac7c19a38841fb80589280d0", + "intent": "Plan structural shard download verification", + "raw_task": "Please help me check this download. Explain how to tell whether a sharded model download is complete and how to recover only what is missing. The download directory contains numbered safetensors shards, an index JSON, and some .incomplete marker files. Base the assessment on shard structure and the index rather than assuming the markers prove corruption; describe a bounded-concurrency recovery plan without downloading anything, and do not treat structural completeness as proof of tensor correctness or model quality.", + "cue": "Explain how to tell whether a sharded model download is complete and how to recover only what is missing.; The download directory contains numbered safetensors shards, an index JSON, and some .incomplete marker files.; Base the assessment on shard structure and the index rather than assuming the markers prove corruption; describe a bounded-concurrency recovery plan without downloading anything, and do not treat structural completeness as proof of tensor correctness or model quality.", + "identity_control": false, + "spans": [ + { + "slot": "OPERATION", + "start": 36, + "end": 141, + "text": "Explain how to tell whether a sharded model download is complete and how to recover only what is missing." + }, + { + "slot": "OBJECT", + "start": 142, + "end": 252, + "text": "The download directory contains numbered safetensors shards, an index JSON, and some .incomplete marker files." + }, + { + "slot": "CONSTRAINTS", + "start": 253, + "end": 521, + "text": "Base the assessment on shard structure and the index rather than assuming the markers prove corruption; describe a bounded-concurrency recovery plan without downloading anything, and do not treat structural completeness as proof of tensor correctness or model quality." + } + ], + "removed_spans": [ + { + "start": 0, + "end": 36, + "text": "Please help me check this download. ", + "reason": "politeness", + "detail": "Polite request opening." + }, + { + "start": 141, + "end": 142, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + }, + { + "start": 252, + "end": 253, + "text": " ", + "reason": "redundant framing", + "detail": "Original inter-slot separator replaced by mandated semicolon-space serialization; no semantic information removed." + } + ], + "constraints_retained": [ + "Base the assessment on shard structure and the index rather than assuming the markers prove corruption; describe a bounded-concurrency recovery plan without downloading anything, and do not treat structural completeness as proof of tensor correctness or model quality." + ], + "counts": { + "raw_characters": 521, + "cue_characters": 487, + "raw_whitespace_tokens": 78, + "cue_whitespace_tokens": 72 + }, + "raw_sha256": "df71e136fa0d3e8acb447155cf2bafb02abd5037d7bf9408377f42658e9568b9", + "cue_sha256": "dca21e95755b0da6f3169a7f71fa27adf943834ea2b8d98fef2df4543cbb3902", + "compression_scope": "Only polite framing removed and original slot whitespace replaced; all substantive text retained." + } + ], + "audit_status": "Independent intent/source audit pending; author will not read source labels or alter inputs after catalog/result exposure." +} diff --git a/evidence/indexer-255-covered-cue-20261030/plan/source-conditioned-author.json.sha256 b/evidence/indexer-255-covered-cue-20261030/plan/source-conditioned-author.json.sha256 new file mode 100644 index 000000000..01be46316 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/source-conditioned-author.json.sha256 @@ -0,0 +1 @@ +e8a52fb88791aa5a1cbe079d9629540dcd104894185955370bd3d65cb842b568 source-conditioned-author.json diff --git a/evidence/indexer-255-covered-cue-20261030/plan/source-pins.json b/evidence/indexer-255-covered-cue-20261030/plan/source-pins.json new file mode 100644 index 000000000..fa60eb419 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/plan/source-pins.json @@ -0,0 +1,833 @@ +{ + "commit": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "files": { + "src/inferrsi/__init__.py": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47", + "src/inferrsi/__main__.py": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d", + "src/inferrsi/_skill_selection.py": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440", + "src/inferrsi/arguments.py": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d", + "src/inferrsi/catalog.py": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5", + "src/inferrsi/claim.py": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1", + "src/inferrsi/cli.py": "0bb83bce32171074506a15ea3ea0617456ad1477b53fd568dafb449f88877591", + "src/inferrsi/contracts.py": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490", + "src/inferrsi/data/knowledge_packs/README.md": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9", + "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414", + "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad", + "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b", + "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5", + "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1", + "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99", + "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8", + "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d", + "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4", + "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45", + "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0", + "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028", + "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c", + "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a", + "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7", + "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593", + "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f", + "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315", + "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102", + "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17", + "src/inferrsi/data/repo_packs/afd_plugin/pack.json": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/README.md": "02394089cbb8dc02c6fbcf96d0c749fcbe5342f5a9df3aaecd73c748e51ce93b", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": "ebecae74a677bae9666a93ed2be828db2abd742a8065af2be5d8ad4dafe181c1", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": "dc2979b9a02ec46937dcb10aeb7e814206b022d5d41612797f428221a55a44d6", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": "269814a1fcb21d70d57f91a05d4a7cda8ae6da4d0849cea065212f9eb60e9fca", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": "6ba2b0f2948a2d691ad8898ae8861c7f6d05dff751ec773d1f69a2914786fb12", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": "e7585f42ab83508897536c17b452c474c90bac14c483b67f1b40df04b93eaac3", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": "e5c19ac97ae2306e2cfd0fd8e035f86d0eff77158d6326e906074da06fac3bc1", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": "3dd78009d21f4063dfb180e1a018270700dd86611bf2f6ae0454e8db06ce2930", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": "ad76590acaf5eee3e7e9b8c1cd8787c7287361dbf0adf0a4a2f818813e6a1736", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": "a067900ca10abd1088a32175edbac375e3141429dd8731fab93cdf49f2aa5d0c", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-video-diffusion-optimization.json": "e66fecf4ff3eeda95a0f6598c5f34de895084841dff97e0d5c16761b92e43d7a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": "32ac2c69ca949d9107e4af738aa49929bf42322bb04481aa8096870f32b18034", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": "f663a87d179efec7e9c987de4d0a7f7d85b677ee00013958b459c76f16082ba6", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": "91c8799ad5b9bf15467a21bc66cf140a6a936613bbeed117cb568e665da78d01", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": "7beabe081115a244f145f1400e86d93733a8fe0274fe0551e0da47a2979bd073", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": "02521aad3508c4e770f02100fc9a19cde48ba53a04d94a24133066761ceca412", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": "fa952d45549bcfd5869a82c85ca55403c877696a1c35576023d202b18d6122d7", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-fixed-workload-claim.json": "e8f96ff65d9a8343330f879128f9e6bdefe96ad348257e9c9e8e7a34f60c6b46", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-metal-fasth3-semantics.json": "98843783777d096c6d35a6d31dc4dc164691d6cfc659d44a485133a88efdcb84", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": "72f25d702005b7f79febdb119a3ff8337d91131693fa6b66ca4126d693800f8a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": "0369527294ae1d21d7686198c4f007137de73daa8fa1cd8bd40c5d0b7ac9296e", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-recurrent-kv-finalization.json": "f7b4134d90d0241bb437acb69c0da5e2467e95d53b580186ecfa1d8e26ef2dc9", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": "02980bf89ad6012dc7c2300a0af0f2ac9841bf918c44000e6df892e01bda0507", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": "0f5914764b8e8a937e2862e040129a8638417f4a336ba52dd6f68e6bec668fd0", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": "3338592a8ce4540ae44e61b28ae57294e42603ef8b99bb8e0a057c1dc688bbe9", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": "987dd19dee9866ea2eba2c2bd661f1be70d87def9400bab37a492ea8df7e33f0", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": "6988ec46905e5c59235718a903543da2129bb279a3b54ea962850c980f23ab4a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": "989525da10cb1c8a8f724de7c52649c877cc2fbca2ad445288db5740346e9836", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": "12d29714fdb78de26571b1b313d2c08f742ae5807708b3513ce7496080e8c848", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": "f35156bb2eef14421a956cbcf9e1215a2e389db6e29edf53e7fdcf186103d8fb", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": "b2d6c847b1cf59e2c2c297b944157d7ed7bb21344988776945994845a83730ff", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": "3f949d221621f6a6385c057281a4215bc9997dfc9eb640d6134e95a6f5d53981", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": "b5e044419b7dcc0022b19c544112485e7de6a5658b7e161dcf9868b2870aea2c", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": "c5c237d0006bb38dccbc5feeb5b5c425668b9c4a5d33137ea34ba5a9bb1e581f", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": "c75c439125b26cde8d9c61cf964dcc6fe4a8eb3eb9590b5d3aa2bb4ad974cdb6", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": "5cbbfb3884b58e900719ef7a30e1e7d17e65903d95000f7343bd5aee323ab80a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": "49b4d4d40c74720e8cb1196a5a742b5f9c280bbd61e16df3f31287f583a067cb", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": "09257528fa037626409961fa96bdab372b5ed060073cd6bc284283e269176f5f", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": "6b5afd25a1f0a37c828180f31098af0a068c9195933ea2d4e77e752af48a1b05", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": "ddf283995dd58337e4805a34266b16cb0f83530d72de39e08c56e98d132a1a82", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dwarfstar.json": "7c32fa29e9bae329da27a2b99659debf01342a5efa234ff1cd955233c78ed062", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": "22b88dd244222ef02484e85e4f84462d0d815a330981cc804605bb1222ada55a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": "941bc2034b553673953c69bfe86259ef9ae981fe010d8e043f7003648d8dee3f", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": "38f9c8b423f0c6a81d3b89fef479ea1689163b7ad22218b4cf67d7cebd568607", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": "fb9413782af4d5bad5c98e72937958cad667e7cea52236672de88dc6def2f6b6", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/husky.json": "fa0742f91758526b24ee8e6d8e34d03eef49600ee2ea24439a0eeaaaa94ff9c4", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": "6c28676ac41066ac2adf2d636b2ceb6514fcb319fbef106a154f5e20235f8659", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": "cdcb6d434e2178965e050d83fd08674440380cf933b4177bcf607d61c2a9eb08", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": "262af0d275ca506fe17960788cae0385f337afdeca5156ee475285c50c2e7ddc", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": "3beff9e39c92798a088742f15333ed9b1bc17b6e274cfe5d15b7a6d079acf3de", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": "fdf10896ead2f2d1b9a563ecf800d6ea6cf554f5134087ef4775aadde578eaa3", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lily.json": "c18a1b798c20859900aef496ef55fb5154bbe751218c8a9eb5400b096b4409ba", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lithos-metal.json": "efffb15e7d008cc61c700d242f3dabf1cbc271a8eb8fb77487ceebc5d6a6ac31", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/local-dream.json": "e03e426f7c1cc32745dd4fe37ca4d93ee7e22667002a87da746ab9632af2f316", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/mistral-rs.json": "16aa560a7e4eb3268b2d75c469a5ab1d45b0bc23e2aa7ddb557280855dd1d9e6", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/mlc-llm.json": "c5a431019f8fc55c0746b81c0eb1d0450d71e62883c9c6d2da029a958e5ca01a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/mlx-swift-lm.json": "98b397fbaede93759c14be4fb9653da383fcd01fbc4c456c5c50abbced0454ba", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": "626d6b934c4923c92d5f98ddf5d32e3f7c33a0526cc3154ab1fe357c33c11574", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": "af4b2abdc0b89bc5a38f79e082e97f1e537cd6648ee4e8379652274568ab5e59", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": "595c86452b42eb91607794882ceee92477d20cce3aa9cd660a8ff075876915f4", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": "24f7d2139809abd77311b99be82ad5ee78da3911a2b2f9a12be9e842e5275b53", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": "f4699d0fbc52b91fc028140b3454b99de38a89e6028afe8bf3c8c0dc30f1a09a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/project-maya.json": "4e77bda9eaa7f8a823351bf10fcdf7f09d8b07714c46b2264883c72e320aeae3", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": "6c218d14c37c15a0f319fecbdc5096d186aab8a70e258cfd3a1657c6dc38798e", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": "31d55ced4a2a4f3d5ff4497f532ba653ad1274f69ece0b1c00d83e926465ddfd", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": "dc68e6a1259677958ae9b0de5c95844a67a60dc6c4b312d209286ad8a3f9d192", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/splash.json": "6f98c43a3b7ea4ed83f5495bdb35b950372f37d0ef53cf0eedabe437b57f1226", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": "529f800b16219ca8b4ad78d8d2122bb8f2765974f7edbe5d1026b4210ec89c63", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": "e1f2f4117096a9c39a01dc15cdf39b5ea61fa422b473e643049e0fc4cb175e8e", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": "f5062359083c6cb004947c0c3f6e1bd52450ec3e92252454b5a399e0a0f292b2", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/video-diffusion-optimization.json": "f87459d7f207d17a071a2525a3463b6d9e911f7157762fbe437959884c051dcc", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/vllm-metal.json": "2e68f1a07220c637c6fa858bbb664a661e93888f65646f2da2809f5df1be1be7", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": "e2bdcb6e10e168e8ad4c1dec571b615581594daf33f7358db1f6b97ddc745b31", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": "d8ffecf654576c0f6470c5cc9f68e8f2e65d494890e23f4c3220ec1212b8e6ea", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-flare-session-lifecycle.json": "1c7da3acaf1f7702a76a790c0ec4fd5132cdbc8625b7afca684aac7165e87226", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": "7f0135ef7920bb50fbe3c5d97fdeb2d2570674ade8245d302ff54a1490ea60a7", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": "68fe4623c829762376dfb789aae14e1dd22f01caeede33750d82d6b235b815af", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": "b9276d6e621be4b5bd9e63c887b2fa9f6cf559b1b54f32f04cd270b8c7e4ba80", + "src/inferrsi/data/repo_packs/sglang/knowledge/README.md": "682e1673fa96c4424f8e1c203e1ef3af174090df4b0f4b00a08458cc835af050", + "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3", + "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd", + "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6", + "src/inferrsi/data/repo_packs/sglang/pack.json": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce", + "src/inferrsi/data/repo_packs/sglang_omni/knowledge/README.md": "7cf067a6162ad39efa42c485c2c41eefaf417883bd7e4039e550c73b28448f7f", + "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114", + "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6", + "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82", + "src/inferrsi/data/repo_packs/sglang_omni/pack.json": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96", + "src/inferrsi/data/repo_packs/strata/knowledge/README.md": "b94b6accf0a2333b751f234d7c60c882aa02a25fb9271c5f75905584ccae86ce", + "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e", + "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9", + "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41", + "src/inferrsi/data/repo_packs/strata/pack.json": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3", + "src/inferrsi/data/repo_packs/vllm/knowledge/README.md": "25bbde60f83028b9aa8647b8c5dbe30107a6a4531a2463fece99c4a5bb585774", + "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37", + "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46", + "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd", + "src/inferrsi/data/repo_packs/vllm/pack.json": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/README.md": "ab8234fe1d420b072038118fdeabdccdf62a13274d44ed58d79c53db745bd412", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93", + "src/inferrsi/data/repo_packs/vllm_omni/pack.json": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5", + "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7", + "src/inferrsi/data/research_profiles/minimax-h3.json": "c5f0fa7871710d80a55120724dbe95544054276e03de65508d6d6e5efdd446b6", + "src/inferrsi/data/skill_library/README.md": "74bd6c7f0201c5296530e263e1e66b0e1dfd231558eb8b6c01a4929b886fec2b", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff", + "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845", + "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35", + "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7", + "src/inferrsi/data/skill_library/hardware/README.md": "ac71975625c55d34491040084131ddc0420942778ba824fcf99c41bffa42e12b", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": "0b79d95e8d5abf0a822f8ae6bfa3779cbb40595c5b6f0fb934233275f5899ef8", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": "56ae2092fb4d6538603c1d3d3916f0a3f0eb443d23dad3217ebd74d48d208d94", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/hccl-LICENSE.md": "05205a02acebd671c78d3b876c1f6097b84cce89780f5d9bdc204bfbea92d0c1", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/hccl-allreduce-readset.json": "4a3d403986dfceaa28605f2f30b79545e1efdc27f35c88755402b75b4d13d7e2", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/hccl-allreduce.md": "4770bba0090ed5ead3d2fd3b2cce2d21ba708acae3e0b4d859cc503499da5b93", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": "a7341b5bf3b43f32d6114e2109e5043e46f3ddeef9b7bb3b20c1c50236bab638", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": "3912bdfc721a703580a97a42ca13ac708396b17c4b600bf9e76375dd061fd29c", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/references/legacy-source-notes.md": "a51c655aff59741cd7421277bd71dffeb9907560950459ae0a776207e496de4a", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/references/torchair-LICENSE.md": "bc07d81fa332b6543463588a120518c57b5ce8582d3747de75d2f6b09b068c01", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/references/torchair-graph-path.md": "272424571ff439278c8f2675406eae7da9a6482c581973e24319739fa0e58348", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/references/torchair-graph-readset.json": "eb84cbd5c355a88ee013dc7e428b274162e0e13e85845b58526113b0d6a2b46b", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-video-diffusion-optimization/SKILL.md": "53c21e64f0dd47f9510faef1f2c01e1283899787ba7ebfe70cdabd8e57a76f00", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-video-diffusion-optimization/metadata.json": "8c2b364298330751fd1252516159b16c27e90354e2a724782d2851e68e43cdb3", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-video-diffusion-optimization/references/field-notes-video-diffusion.md": "dadac7b23462fdc325218c72ddf8a2ac553151761a5d3e45b854b53d2c8c6908", + "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b", + "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71", + "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2", + "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f", + "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d", + "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9", + "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5", + "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": "4c8db9f603cd581d002e8f41da65af9cbd03699be05c4965a55de47e044e4dd8", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": "0d1fe7238bf3e744b463df5f7dc683110468ae266709e63f0d3edfaa0e2f110c", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/cpython-PSF-LICENSE.md": "33f1a470702f459c160751bf3188b3fa7462fc761a006e03e80ef3f0a7a01ed6", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/cpython-free-threading-readset.json": "46336623b3c573fad97d4c686d9836baab807adf10f4bea05352d7a46236e004", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/legacy-source-notes.md": "bb1c0fb7a4d1a0301ca9e699e733224f3dcf0ded3cfe13e612e5c04d48a8cab5", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/pytorch-LICENSE.md": "a8a2c6b67cf8a597ffaa5c82380a6bd1b82629c0417e00673d8013d662eb500a", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/pytorch-host-copy-readset.json": "881258c97edfee0a89f2392c2ff61cd7fe37ff8344f18620f281e841fb92cbc8", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/pytorch-host-copy.md": "fa9da5c39d37ebe44e5a336a7b88d7aece6a0f118ec47cb38271e29925e11117", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542", + "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb", + "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a", + "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924", + "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053", + "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5", + "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e", + "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7", + "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5", + "src/inferrsi/data/skill_library/hardware/source-inventory.json": "9e8ca783d8d9fa35fa80b3c045584bdf42578c81767da5c7703b15c901a2468d", + "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d", + "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0", + "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f", + "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3", + "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7", + "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e", + "src/inferrsi/data/skill_library/methods/audit-fixed-workload-claim/SKILL.md": "2be89b1340a3deb978c8e1792d18aeecbe8f5de81d4daf542389f53826b8ebee", + "src/inferrsi/data/skill_library/methods/audit-fixed-workload-claim/metadata.json": "81edd75fa16939807548d0944e9203cfd7c0cb17e63463872d96ca2c1273d1d8", + "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352", + "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0", + "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5", + "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688", + "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328", + "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc", + "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233", + "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24", + "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7", + "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b", + "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933", + "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490", + "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc", + "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5", + "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b", + "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": "59b5656e7d989b3cee8d6f95cd43be03053b1fb11c3a82bb49ab3d402f7aacee", + "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306", + "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3", + "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": "360361f592ec3a24834198372cbf92aaedc1720404d182df94315eb0e01d58dd", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/causal-conv1d-dispatch-state.md": "2eb67dd522021bb2234616d7f660e6ba99acb43676f48cb23141bc69bf3e0029", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/hf-qwen35-gdn-binding.md": "b3a55af3f45f01c48e1cb18c8b029e5949a494c3e51dc444a5cccbad9acbba0a", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/sglang-gdn-source-contract.json": "6c2586ec80698dba848ed09024954d060307be71c8701719af19df3b735cf595", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/sglang-gdn-source-contract.md": "8b093faf0ab775d5605ecdd9653fb3689c2b839cba2641c5d5cda17c721bb2ac", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/vllm-gdn-causal-conv-contract.md": "2f3dcee66e2dfe7a98bee21a36b892538998de5de3caf1bc63257137f9984236", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/vllm-gdn-source-contract.json": "73f1a2d8fa89ce05aead2091a05a285b13cbdc314d05452c35c6a2ad15ffeff7", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/vllm-gdn-source-contract.md": "7f9ebe8de7e55d3db2a119b6a56e15c37448fd7d4e022a47122a143206d31687", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": "c86f93f4eacde789e5438327d85b1de0bfe623fb87acad49306e47b7e8ef1261", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/causal-conv1d-contract.md": "b3a6245cf5e9deeba0dffd4a15ddcf97da6a4315f4d3131921779438832663ad", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/hf-qwen35-gdn-binding.md": "848d3c99b47ef711e8820bbc4f6bcbd1db7211197811c9d855201025448d806b", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/hf-qwen35-gdn-source.json": "df3f4650b5cf92d4fd3d5a4cc8aefb91dd4b0c2affaa0c873138d5651c50e6bd", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/licenses/sglang-LICENSE.md": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/pplx-gdn-caller-context.json": "0219fc9ad785da4e638c1680bc9eaf28672aa32313824f52bdcaed61d29bca87", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/sglang-gdn-causal-conv-contract.md": "c9f27b0231ddec3d14d6fa0831d1494ffd8e11bd8ad704e04be3bc4a4ffc0e58", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/sglang-gdn-source-files.json": "4c7ec2cfd2c29297e185ce09b4d9b2040fbe3635943c112f9ebb71dca8ab381b", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/vllm-gdn-causal-conv-contract.md": "eec09a6e95983c46b67d5f066d34419c17819899f5d8e3c65f2f163f2ada0fde", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": "dd275a0576e827353963aae7f0b09494d60d0b9a4a2884fe575c85249fa3131b", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": "4a61aab6b9984dacdccd657c17f2d959b3a6ffe74bfd2f1677438f3bb8f025d3", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4", + "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e", + "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7", + "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3", + "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": "61bc684d67d6c0876c4956057f11c88802cfc4b771553d00411634d8ea226b3d", + "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": "98ef753a0e74c46f59a69ae267bdc0a5d293761e6f27edc1822fa8bd4548c4d6", + "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5", + "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/rtk-consumer-path-transition.md": "a70b1d128fb5e0ac12292a290d624f218a0831d7581fe8980b62661473e0f2b7", + "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3", + "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8", + "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e", + "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac", + "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f", + "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef", + "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744", + "src/inferrsi/data/skill_library/methods/video-diffusion-optimization/SKILL.md": "fc0e7baf9913ce7448247772a1e224a9cbd827492cf9459b1f17f492b9f67b70", + "src/inferrsi/data/skill_library/methods/video-diffusion-optimization/assets/campaign.template.json": "1a37f1d136612143ddab8f0cbbf424c80433b4da5312edb18d07c0b9d87ef657", + "src/inferrsi/data/skill_library/methods/video-diffusion-optimization/metadata.json": "b27d69142dfea5dd2820170475c2bf6cced2d5eefd1c43b3413c41849af0fd02", + "src/inferrsi/data/skill_library/methods/video-diffusion-optimization/scripts/evaluate_target.py": "9e84df051e03d567746fd73ae7243efb2f3c5ec806fef824d3c4d42a8eb9704c", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": "a1bf0e2ec066d5f30225c6cc5fb603f830f48eb5e2aee3fc3d084ed9d8aebe2f", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": "6df47ea943e42a11807c54869774957ac2383266120caad730f788207403581c", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/flare-preparation-boundary.md": "1d7922785a196bdbafdc9791fe79428465a5acc0292ca8f99446f8c764b35454", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97", + "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535", + "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1", + "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943", + "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72", + "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986", + "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e", + "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841", + "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac", + "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7", + "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e", + "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263", + "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5", + "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131", + "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098", + "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff", + "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307", + "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919", + "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6", + "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288", + "src/inferrsi/data/skill_library/repositories/dwarfstar/LICENSE": "4a8a1358f5b048627d71e4988de0bafd386f4988991929629e6cb9e29b7b9b2f", + "src/inferrsi/data/skill_library/repositories/dwarfstar/SKILL.md": "a04a252ecdcd4710a3273595f4bd1f3eb8e24d5bb1bfe005f058a78388254dc9", + "src/inferrsi/data/skill_library/repositories/dwarfstar/metadata.json": "e1f0c2dc3d7ea9278039c2e8c3e8ca4f8e3aba33504b7784d628e903f0b5fc8c", + "src/inferrsi/data/skill_library/repositories/dwarfstar/references/repo-provenance.md": "7d72f9c85ee60fe003f5f5a585b52e9989ebcedca625b584fa39a976d4433998", + "src/inferrsi/data/skill_library/repositories/dwarfstar/references/repo-routing-metadata.json": "8e195b420b67151932af5f435c284f49677cc7863e650c83cfa9e762fd00db0b", + "src/inferrsi/data/skill_library/repositories/dwarfstar/references/source-identity.json": "a38fe07d003accd83b7626d7e189f7b49b4b73e7a223e4d441b0e33add5d9c91", + "src/inferrsi/data/skill_library/repositories/dwarfstar/scripts/request_template.py": "62e4cde002f2aebeaddd63e6e1f04fc727c271d2cbdaa00856158baadee5340b", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/qualification/SKILL.md": "3b8eb25f6b8e4729b48cce5619daa24422bee251f119ea7f9cc1faf0097ec6bb", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/qualification/references/troubleshooting.md": "cd832ab2a420490524251a042c0c83c2f75bbfdc12e914b91da8f91edf6860e9", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/qualification/references/verification-contract.md": "2801f6683c2974d1cfa5b86c6268fd33f175f1cbcc9ed85c932e9b3504fac3e1", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/runtime-memory/SKILL.md": "ad1d34d429baaa41fb72c94ff4e554b26b2f5b701cc4deed03f1d066c7d3fc52", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/runtime-memory/references/memory-ledger.md": "765915ecd3a318cc7908658d5a79bbdff615fc24409501febcb7ff8af65a1050", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/runtime-memory/references/troubleshooting.md": "42e4785b7d3c1375efa324583897327f28ef76c0b7788092da92095bddeb1440", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/serving-state/SKILL.md": "71c27212847146376d72e6cc6d22abe2f7eed56402c5917cd7dfcda0b2aa5982", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/serving-state/references/state-contracts.md": "8f74a07c947021e3069468268f6b5aa2548a17db518c5c0505b84dde73a1f31c", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/serving-state/references/troubleshooting.md": "c9eebe4109d418c4dfb9bb2425c145ebc1a52da9b665e0556149bc31a442e2fc", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/SKILL.md": "02a5f7c18e44201215ecec17c6ac96b27be9148bcc189df287f39d1b372b7721", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/metadata.json": "58716737446f46bd1de77aa323c5652f97277d44dd952c82b42399b2cb769b87", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/references/diagnostics.md": "0908a0eae53a2e10775a8b6ee3e3944b2fa410174402b5a1a8c98eb674dd4b4d", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/references/lifecycle.md": "107ebe3c22e5dba8007c3b8ea61149f10bc55c51b557a4230f3e5de859bb0f7c", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/references/source-ledger.json": "61e51a97ffd7c1fa7866bdd440b024d0c5e0065635ac8d0ba53d2c69c7f9e470", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/references/source-notes.md": "fc7ceb819b91ef1731d8b7783f543c92ef6bb5c97ef5f897c055824f8a560168", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/LICENSE": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/SKILL.md": "e5205de6eba52b3d1693666af755640697717e52ee7834e1ae30e4071b208d54", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/metadata.json": "86a67bb56110c3a9e191e0ff576194162e960cbddca0a99c593ea7aef8df8a4e", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/THIRD_PARTY_NOTICES.md": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/preflight.md": "5cf745005914c74b18a09f127d1d02ce0cb64f6cd7899501ce3cb40938c42a4c", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/repo-provenance.md": "6a4c14207223dcc2584f7133bd14ff4590c2258fc2bd44f3721b6bf63e3e35cf", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/semantic-contracts.md": "3f518160e8a227272e698486e6c67acd364af0d74755103ffc868b2d66052c20", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/source-identity.json": "b0e63197725c6a1baa4888ca3c48ccd4926b6e6f0f8bf781ae76ac97ffd1ceb1", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/source-notes.md": "f9430103644eacd9c55ace6f84d8587fc0bc063bfcd9cb60e6045b139a9fee05", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/troubleshooting.md": "ed552d87f9fb9fb2c56de64df61f3fc6e2333cb049116cd0736f784c8a327612", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/scripts/audit_fasth3.py": "05bfb9348cf88a76f4fba8bc8a600fe54246d83a98c1685bc4847ace28e2fbac", + "src/inferrsi/data/skill_library/repositories/husky/SKILL.md": "ac86afdabf9adffb684cb635971934438724bdfde8628ec549727ea23c40cf3b", + "src/inferrsi/data/skill_library/repositories/husky/metadata.json": "9b720bc6fcb65b6dbddba36d6187b4bee86346a1bc633f55a75bd43dbe4fa04e", + "src/inferrsi/data/skill_library/repositories/husky/references/author-report.md": "4c1351ca42b2535a7c03c4da87706040650ef10d86eadcdb496e881a147fc0f7", + "src/inferrsi/data/skill_library/repositories/husky/references/repo-provenance.md": "37de912e2e2a36b44233560e6a8368b994428bff587ed4f04b1cb5fbaba0a9e7", + "src/inferrsi/data/skill_library/repositories/husky/references/source-availability.json": "d3523e53c89d4aee5bfaa17db84501aa4c4eddd59bcee486df5d48d3aa48173d", + "src/inferrsi/data/skill_library/repositories/husky/references/troubleshooting.md": "3948711437d25822d389dd857c72dc7fcc15727c9048d2e3eae3317d97aca6dd", + "src/inferrsi/data/skill_library/repositories/husky/references/verification-protocol.md": "83c5a966ed3f721b412b8d8417af8336c5a8c1e46044a888720936bf2f1deeab", + "src/inferrsi/data/skill_library/repositories/husky/scripts/audit_evidence_card.py": "291e62701ef7a37a5bfab963ea4e20abb94245bc015584223a101f18b1490895", + "src/inferrsi/data/skill_library/repositories/lily/LICENSE": "b9e2b50a48a09fb39dafa71ab1a11a6e87389addda17faa99d55f5f74a0001ac", + "src/inferrsi/data/skill_library/repositories/lily/NOTICE.md": "7a01e666044fb8aca1fd88c636bca43bcc95b8b1fecbe101dce4c7c80fbb5759", + "src/inferrsi/data/skill_library/repositories/lily/SKILL.md": "db815674d5d36964902041d310d3df844d00ab2e6bef0ce4b72ed2696ab2b9aa", + "src/inferrsi/data/skill_library/repositories/lily/metadata.json": "6d9d54e84f957881cc26894eb82ae666f106cfc6c505924f9a6fb6b35eab457f", + "src/inferrsi/data/skill_library/repositories/lily/references/repo-provenance.md": "96182f6cb3666f89adf1aed6e216c0a84ce3a04a0685a4257e286100ba8ee367", + "src/inferrsi/data/skill_library/repositories/lily/references/repo-routing-metadata.json": "f2f3becb51056e03081c7c4140e19572a6318c3d63015884df56eb49d38a00fc", + "src/inferrsi/data/skill_library/repositories/lily/references/source-identity.json": "4595559c11611c855be150919fca3eeb89e7f96c7a8dd133b9f79a5cd8e05e62", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/model-runtime/SKILL.md": "dc800f07fa0e82213483c7b848dea95d609386296006bfec7a34cbb194a83de6", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/model-runtime/references/model-and-memory.md": "d5a862303d646ebbaf745818caed8f96c7323cd442cb6d6e2aae00b0361080a1", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/model-runtime/references/troubleshooting.md": "b32fab02b561c3ab4e683abdfa80cd4a8f929c514147709febaa6b6d1df65909", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/model-runtime/scripts/screen_config.py": "5961bb6f31ceca43473683127b37faee268234040a013ae49bbb4898eacb6f04", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/qualification/SKILL.md": "0231b50b9bebb68d0fc5f2ce19aa7c6e5ffd498d8eef4cc00d39e82908cbbe91", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/qualification/references/troubleshooting.md": "01607534ce1f09fb046255c78f8c1c436aac61874b93b65a9d62f3eb480496c2", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/qualification/references/verification-contract.md": "6934779f67b878bf7cdb5f2cbe461f6716c243c662a29272c6db37f64cb317c7", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/qualification/scripts/audit_benchmark_record.py": "59534f58d46ad438c7bf88886a9b05a8716c922d817f2d9c58ff1b71ae50ee85", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/serving-cache/SKILL.md": "7e062ff9af76125ef08c98b4465e0439c66e9e237f0c39ea0285005cc5496cfb", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/serving-cache/references/api-and-session.md": "7262bb077c24f2f7549d450473ec7944829940d09cca89db9c12f37ac90d5db0", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/serving-cache/references/troubleshooting.md": "6e2f36d2f7ab9eae665d87a676e34c0ebd2f46f3df6fa5f2e589a306749a87b5", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/serving-cache/scripts/request_template.py": "d6b00c9f0b38c7fb0a0e52e396993a832542ea0aa6e2456026fabc0c39a4d33a", + "src/inferrsi/data/skill_library/repositories/lithos-metal/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/repositories/lithos-metal/SKILL.md": "4c71ba2f26c275829466bba2412738aa27799842e642c75cdfd36d9ed8713f7f", + "src/inferrsi/data/skill_library/repositories/lithos-metal/metadata.json": "ea2d4e73e33a49433a2834d19489653f828e56f69c4975cc32cf508176eb24ae", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/mlx-license.md": "ccfab7ccb2ea306f71531c8ca77bb55507606cd90768b1e32b8b52ab5b48cf01", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/notices.md": "b2c9d2481ef2c9d16b3240c951ae65db417e5d309cd06b7917db9d84fe175886", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/package-api.md": "4f3ecfabafb2a15eddf21ccf964b8c37635bbb7862418a2c404e7049c0fa2898", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/repo-provenance.md": "c0c7c3dd45eab0d55bcf154760156db58e8e365f9cfb2c6bd246dd6a06f3ef3c", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/source-notes.md": "110dbc995c7fc727632017d4c26c71df7ecf12d4df5d89df0cd4afd75c5892f2", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/troubleshooting.md": "1894bbe1a18018f81318b11e2150201ee97f3819f84a7a81f3cf5fb01549c165", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/upstream-notice.md": "cd80bddd6b6204dbc61c6ac762ed206c73d9264755d9884383d0860de29d185d", + "src/inferrsi/data/skill_library/repositories/lithos-metal/scripts/inspect_contract.py": "f8365f87af2135fcf0a2c8baaefac563c30989282fe080a02d958a37a215515c", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/compile-fusion/SKILL.md": "e0d927763f3774440abddefc45913528946cd29f7f089766fab47b212f9b6693", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/compile-fusion/metadata.json": "b3ff3702531e023b49beded950151ec1256aab4fbf4c02e9212ef0202d3ad72e", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/compile-fusion/references/mechanisms.md": "6173a8b73ec5138cb12791f108f3d5d83278a117d443d48187c842e16bdc6eff", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/compile-fusion/references/troubleshooting.md": "f18ca9324d9a80f2ba327a1ad4ccf84a01049080366fe12410f2800d5f178232", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/generation-state/SKILL.md": "bd8cd27ddb18a04c9e2b31ce9f16e676864133e160267473979de828b1915f63", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/generation-state/metadata.json": "dacba1953fad82b32b98f76bd48009b04109f1ae446cc68be522b02a64306157", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/generation-state/references/state-machine.md": "d812ad31d4e343539119cf82f8ce241fe14c691dd73398048893cc5de3ab8e22", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/generation-state/references/troubleshooting.md": "7ca629e66ef3b0a67077a7d2536befef683315ad01939728a9e6e7523cc8414c", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/serving-cache/SKILL.md": "01c38c2678d6e8292a3908bf5619b332432c5f1289fd5d5a67eb44a2141adb90", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/serving-cache/metadata.json": "2fca222efac7752e29f236f40a0bc38275045826b55955687e39c64ffb69f58f", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/serving-cache/references/troubleshooting.md": "52c6e27a76547475b5c66121f873f4d2c1662c811a69da64ce85614f1bd44749", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/serving-cache/references/workflows.md": "298d27d3c748fe22774b872ed0465b707e53812379a7a538c95f25d1727fc796", + "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013", + "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c", + "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190", + "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27", + "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829", + "src/inferrsi/data/skill_library/repositories/local-dream/SKILL.md": "e51fd3c71d693070abeba57ca67dcca82c7347cb94eb98f43e7c455bcacbaa46", + "src/inferrsi/data/skill_library/repositories/local-dream/metadata.json": "ab33602252d0f5bb0797c3e1f12466b55a82aeaff10cd24e42a99a5224c8e593", + "src/inferrsi/data/skill_library/repositories/local-dream/references/android-lifecycle.md": "f85e5fce39f5de5f88548fd8c21978240c810f899e03ff05ed72eb3508ee5e36", + "src/inferrsi/data/skill_library/repositories/local-dream/references/native-contracts.md": "40206892f5e421f9d8edf303870f5c4aefa3e87123537178d45f438f3d04b7ce", + "src/inferrsi/data/skill_library/repositories/local-dream/references/packaging-and-abi.md": "9a19bca59e19c3bfbfc1a15f5fc756c13d954ace367812aeb000551f4484ad06", + "src/inferrsi/data/skill_library/repositories/local-dream/references/repo-provenance.md": "0b848082c0ce893d8ae26d779c5b663d914d1a208cb2293c4a84668cdde86293", + "src/inferrsi/data/skill_library/repositories/local-dream/references/source-identity.json": "b85f028bb96a37852e9d9780b5d5052fdd9c8f3a704ad391b6381adb58291347", + "src/inferrsi/data/skill_library/repositories/local-dream/references/upstream-license.md": "41003d4a74749c0220e33dd415042164b5a1093ed401f36277234f772d22d3d0", + "src/inferrsi/data/skill_library/repositories/local-dream/scripts/screen_declarations.py": "ccd7d0d55be9a47484ec3a74498d8edef8aba55287f92287775c8d82050c8932", + "src/inferrsi/data/skill_library/repositories/mistral-rs/LICENSE": "b40bb09c7b3b6f697a8920cf9d2cb3ff8cfcf77a0fcb6d264a92842b43c4dcae", + "src/inferrsi/data/skill_library/repositories/mistral-rs/SKILL.md": "8917b30fb86437dea7508275ccce1992194d6e0c4858d537a410bf1f900953d0", + "src/inferrsi/data/skill_library/repositories/mistral-rs/metadata.json": "5e88e2ca64b848b8177336351e03d1fbcb5c495fd54217daa83710ef6faccff0", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/apache-component-license.md": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/paged-component-notice.md": "a1718d8c5a25f15e61bea98a158880e96faa5b7174e82c4e1470fbacb23a3559", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/repo-provenance.md": "fab87dacbe0a7298e42137cab7409f443ece0dfefba60961040031e36d96adc8", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/repo-routing-metadata.json": "0debda77a8f6b1b1c06176e3c2447767c1953b9fe7e4c84ef01b3ac90735df70", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/scope-and-notices.md": "938203995f5e58cd6cefa278225b2d5a3ea7dc7f07cae86201f9fc71553ac020", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/source-identity.json": "6a8c0b0834228b3a298e7e548c9d8cb774c00f661135f51c12ee613efe1decf3", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/SKILL.md": "73620deb4efad4681da10a07bb982e180423bd69fec3572d16ab556c530b89ec", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/references/helper-contract.md": "802d8af65343eeb29b0950d3c305d0f24cb7a4298628ade5f7cdcf4358c9e228", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/references/layout-and-features.md": "6804a771ed0dd2d8d60613d52ef87d4297af2a6445b4fe61d08adeab4be331aa", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/references/troubleshooting.md": "a6de920f174d6918a72093921d344c700108b0c01e1f0d09f2b9051800b1fa7d", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/references/workspace-and-evidence.md": "7ac5028be6fd48f9c6cf067bf832efc8edd681195f2fece67ea119211bdac106", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/scripts/audit_decode_metadata.py": "de28f586f27278fee25f6c96f1e00ee63528cdf746f57c83bdf367c31c0f6e6b", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/SKILL.md": "21015790789933e106f6ea929fd1778d862315520b55425213c679e9b7a44e95", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/references/helper-contract.md": "607449105bf3bcbcfe785dc9448a7a587c7daeb528ca63e0cd959b074a9bc409", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/references/prefill-and-metadata.md": "e8a738a668a23068b8ff1907d521a3a26f7090e7a791f3a715efdb6001338dc0", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/references/troubleshooting.md": "f584a02528e2249f5d12f223d8933e9d40212e5aba8dbe545261164b00ed6f8b", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/scripts/screen_prefill_metadata.py": "769366c19d621ec1b42c722ed224fa154a63c3410a0cffc9072a13abd9c71bbd", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/SKILL.md": "745d88dbf6c764f64f8b31763e57bef720788830ddf17b518456323000325fbb", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/references/prewarm-and-evidence.md": "6b72ae4b4599018ce2da5c54d0c28fadcafb7ce75b81ce4c18c06a440aab6056", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/references/troubleshooting.md": "d933bea6e9e493faa8cd0f4103ea33197ace7f7f901359f8cf0d6bab2edcdb4d", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/references/warmup-inventory.md": "1e7af7d63c6853ede4232b5a12cf68aba6597c4a305015366f1c2c917e3caf01", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/scripts/warmup_shape_inventory.py": "6aa920ca1cb534dec41f5dd2d833c3e760b234ab2c6439b6c4250b8a30b74055", + "src/inferrsi/data/skill_library/repositories/mlc-llm/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/repositories/mlc-llm/SKILL.md": "e0fd1accb4216191c56e7e4a8074633f36d9063727c2546e0d2eed14220d8cf4", + "src/inferrsi/data/skill_library/repositories/mlc-llm/metadata.json": "c9927e284e44aa09733424c024409ec8efc9593520418bb7d2e53ef301363681", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/mlc-notice.md": "3a7dfb37bb96ce6f5a1cc0d6d4c8da3b8a35e6ec4692647ba0a37f6155c7e694", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/repo-provenance.md": "423f21e1de6d7ab2e11c0bdf540357efde025ae4535bdf76ed9770b1a59ba35a", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/repo-routing-metadata.json": "5309c27e7b60c4fe7aaef28bf6911e7e3c936b58b1a907aa8486bf0e40b4c0d0", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/scope-and-notices.md": "8c573c234f44a0808a731774921fb5228fc0844cd238496a2c4d61d0120bdc79", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/source-identity.json": "75e91b0f7fdac12466e62eedfb49baa6f1b3016111d7479aa2b04e5c799454b2", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/tvm-license.md": "e62264b483aca7a9f8c1ac69388b476971b43975ced90ac11dbd9e9c530edb46", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/tvm-notice.md": "967cd2fb97712c43271e2b1c4f636c4d8823898869f0f76f65bf8a3232584f0f", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/SKILL.md": "933fdd05ca853f47e3ace024cf1e08df7511834c31676851c28803b6dcb1c681", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/references/contract-mechanisms.md": "82187f4a1478352c68a5690101e9798973ef63f9b6d717e06188b03a1832bb8c", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/references/helper-contract.md": "48492a36435ff1340a4db4e0959f052cd72683ced26bb8ab35baa84bd79ccd7b", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/references/troubleshooting.md": "41b002a31b031a78ca372504de4c9cccc9353d6d80d989b8d62e56260c291e2e", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/scripts/audit_artifact_declarations.py": "874a16cb16d75a14cb51b4fbe442f4fac46feb21ac242063f5f845a0fcbabaa4", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/SKILL.md": "25dbd48f4fe84b7d759005d04a2aa9db33da987cafac7975a148e62cc9efae34", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/helper-contract.md": "5428efd6dca0654a58e9e226805e15c9a9f1b3f1ef60db234139985d107acbfa", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/modules-and-streams.md": "56faaf149f95ee821c242ccca8115a6f3b166d34392499798ae8532fcd10372d", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/native-evidence.md": "133322e879fd9010a7513afd278d937838aa2ec97d593ed11d324cc59793374e", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/targets-and-export.md": "547eb8e56af4f7b58415eb58c338ea1a33ff8edf8b58fa523700947dc7e04821", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/troubleshooting.md": "e392b8329ed183bba5fd79fbd260c3fbfb389016ee71880a957741e71a0d59c0", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/scripts/audit_apple_target_declarations.py": "d5bf16f7def4cfe28b143a700ca45a9636ddaeac12ead04cbaecdcc8b4b71aa5", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/SKILL.md": "6b4914cb60831c07d4c748162251c49fc16b229e0d62b4c5a5663719c45b16de", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/references/cache-lowering-and-evidence.md": "a118f5bc584ea885e8c573978eaf1e32d448910860b8135196c5c1acd42aee05", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/references/helper-contract.md": "da33306472cbda5a33605b80fe52d6cb1fa886b3b7eea13adbdf0ce8a6024c66", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/references/limits-and-workspace.md": "77105db95de2b8ea27ed6953b5441c8af061fb745ddb13d3904f8773fc21e879", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/references/troubleshooting.md": "b89255fc04507185bc8bcf74de779730b009b12d13bb38ad0193205dfed8e921", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/scripts/audit_kv_limit_declarations.py": "d82a6dda5f5931eb369003a5e1f419bfbdd82304766a18be8ea1e1849e26b77d", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/LICENSE": "7b1d86acac816ce8192209d7e01141c129b4d48d474c15cc5fb4af62f497c323", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/SKILL.md": "501f73e7b327b04b6b80501b42b20dab1ac15c875817e6f4c4b66fe3e69b4fc4", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/metadata.json": "014131f345a28d3367313ed5a4cce858fe4c93d4d47a5f03eb89f609592e032f", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/references/repo-provenance.md": "f1b15edf89c597db0c3fc0bff7db6b2451924ce9273c39623bc0cbdccb568d31", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/references/repo-routing-metadata.json": "8fc804347771cd768a7ed027f2041d96c9832fe5f8c5869e6593a3021b1bef8b", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/references/scope-and-notices.md": "18e56805a8f275d35af3d8e42f3aa1525871b43400921a5d2dca4897c2047214", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/references/source-identity.json": "518da7bb64f22ff8ee6beb0e0f0579e7cd0da2fd70fa69fd8288294b0c08ba24", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/SKILL.md": "04b97959b4e207380a06d932e9a557954a9fb9aa880c243439da6d4a981854ed", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/references/helper-contract.md": "1358531257e00c0f9fb2eb9fb7a40f9d749f5435c90168c8923b6646353dce6f", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/references/recurrence-and-state.md": "ff0ebbeacb8219d93a890418d18bdf2856e6326d67f974a7050de3a0e5a3c5a3", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/references/troubleshooting.md": "3bbecbf91b241e7b37b3e8c753bf3fbcbdbba0b0f4d0ddf2f65eb653007f6cff", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/references/verification-boundaries.md": "21b6fceb692f29b410dc967f82e31d0d082904cea827443ffc05f6f940191dfe", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/scripts/audit_gated_delta_declarations.py": "3a256c37c867866e49dfdc2893cbb3f32d061e389fe4395555eff899bc0415ff", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/SKILL.md": "7effaa20ddaf32e3b4008c95888ccd09bc2bbc3bb5ff17c710f989b36be49535", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/references/evidence-schema.md": "d40e44910d8fc43ee3c731735902b1c1ba506c2ae972bb53dbacfaf37e38509e", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/references/native-evidence.md": "3e9a630a30932989faea5f52b347318cb94c73c7d4d6c81c0465ccaee2818671", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/references/ownership-and-termination.md": "56fdec7ebc89b5c217c7e5c203886bef1d3c2eb5794852897bdb5941ab1db2a0", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/references/troubleshooting.md": "de28b8ca9cf95be2a4e5762f938acfadb6d4b53459dbf86df4b2b2ec64ec347f", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/scripts/audit_generation_evidence.py": "2aba6caf42ca1b3fc55299fd96419af8681c37750c21a7ba75205a43baa9c7b1", + "src/inferrsi/data/skill_library/repositories/project-maya/LICENSE": "574cffb7c57769a9271043ac0c2b07c5e7e7621cfb81081b901d8981e791d121", + "src/inferrsi/data/skill_library/repositories/project-maya/SKILL.md": "c619a3472be02147a847a24ca97c56afd8b56b6ffd22ce66fcae453155e34bab", + "src/inferrsi/data/skill_library/repositories/project-maya/metadata.json": "c9f7879ca512c07e9bdf6d1d1a52de055a0fa210b8838c205e27a34436324834", + "src/inferrsi/data/skill_library/repositories/project-maya/references/license-boundaries.md": "090839df90d6e96f55794c2ad821d7154dc6a6aa4211cdbb08a1e5f6f2e52a92", + "src/inferrsi/data/skill_library/repositories/project-maya/references/repo-provenance.md": "7bb6b6c14d1a9b77e906804505c723569f952e5d47559e636d2e6621994216bf", + "src/inferrsi/data/skill_library/repositories/project-maya/references/repo-routing-metadata.json": "0acbb111e92316e74d537200cd033f3d7d0f69972bf0953c8cea9dfb2ecb4a34", + "src/inferrsi/data/skill_library/repositories/project-maya/references/source-identity.json": "fc2287d5bde587cf040bbf515161d25e84a7bb42989cf01343582af67e4b8eeb", + "src/inferrsi/data/skill_library/repositories/project-maya/references/troubleshooting.md": "735d4bf3a3b243c29e6371af37dc06c1fa43b39a85811ba9714b911bf520ff67", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/SKILL.md": "1b68800cc2dd9b7b03cee3343fb8619135dc3563550cc20386606c295358df6f", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/references/model-and-memory.md": "ea7b905fa72cbfda5513f4c6ae77c552044af71e9313e396bcd3628884183bfb", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/references/notices.md": "6f6d9a96e1e67da6fe25f54482da0b69bd256fb688cc0956668f7319308f310e", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/references/troubleshooting.md": "97c00435d57a5b9067fe8e5885219d5818208a25f225a14c240d7160dcd697f7", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/scripts/audit_runtime_plan.py": "b096ec10c4f0e4b057997fe7b946007ab4c06dd263ab09d803e29e498601da00", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/qualification/SKILL.md": "5e264000198a9bbdded9c35db2df682e8eee4408ab741f18bc002e9be103b401", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/qualification/references/troubleshooting.md": "0c42256af4869e5cd1eb88fdaacfe65adf51dd180bdc03a2b9fccf220a19ccd8", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/qualification/references/verification-contract.md": "eaeb9319df27b66840cdfcba5c4788a287775d1493308517d9124fd0608eefec", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/qualification/scripts/summarize_decode_records.py": "cc0ac3c403aba79a131f631290d68e1640283e9f0d564aeef431b577824e3036", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/SKILL.md": "830d8bccf6b36e1ad6cef899d5147afb74c0514516c68d02060bdfb550b56920", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/references/api-and-history.md": "a51fa298017f266decd801441eb411fd68f7906a90a4b0d18083d1af31fc7d52", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/references/native-state.md": "e5a7ee5e777998db0fb5422cf5b1ab18dbd67b4cbbfcfd6628ed6e40d8709c61", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/references/troubleshooting.md": "97dea7b2d26fe165d2f64a1a63c91d099dddb2d3b95c16b667f5173b3ffa9d2d", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/scripts/request_template.py": "d91b03fbbd7003012a92bd7259d39604b9a91ee5707da2d3366e7c0fb47edd6a", + "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015", + "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c", + "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b", + "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c", + "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f", + "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747", + "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": "f2b68d4cac90b5bc4d0e38cd17b9f5bec862e934668a459ee08f69a21b5fb72d", + "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747", + "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2", + "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6", + "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d", + "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7", + "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93", + "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b", + "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38", + "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c", + "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794", + "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c", + "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f", + "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e", + "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc", + "src/inferrsi/data/skill_library/repositories/splash/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/repositories/splash/SKILL.md": "619b87c282dcec4befa319085764fb53fa506a05c6a7112500bcca662f059ece", + "src/inferrsi/data/skill_library/repositories/splash/THIRD_PARTY_NOTICES": "2fedf1277bf732f8435c3e3be2798f52f25f7e2bd4155865cb366da55b88e50b", + "src/inferrsi/data/skill_library/repositories/splash/metadata.json": "d4c679f474c31549156aad417e23c2802b9f4bc904472e5cb16916f2996b50ba", + "src/inferrsi/data/skill_library/repositories/splash/references/repo-provenance.md": "20e3462f1b6a8752eb0fa6f36e76867f142a49c1953f498f895e29c4fd4f0204", + "src/inferrsi/data/skill_library/repositories/splash/references/repo-routing-metadata.json": "287ae591f42949b8d33312659ec8303c47d5c73093c64c125edb1cf6ea8de8b9", + "src/inferrsi/data/skill_library/repositories/splash/references/source-identity.json": "2f778882ec440538a46f48332650412b446722246c8dc5690083f19da7dce61d", + "src/inferrsi/data/skill_library/repositories/splash/scripts/request_template.py": "086d773d42edf67c43cedae16974061b09ac66458ee47fe961fc01733f3b2a0e", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/qualification/SKILL.md": "4fddc39811c12ec4b3279ad68a8864b7f1ef80e838ea03788b54cb025a03ae16", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/qualification/references/troubleshooting.md": "dbd2179a48eb3fc0ecfd54c5df8a690d0cb2e1814acc19ac3b0aabf184092c92", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/qualification/references/verification-contract.md": "9efffa0ecfd3b6771dfce631137eccc6d939c3239f60b1fadcd0c6c1ed7700be", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/qualification/scripts/summarize_rounds.py": "107a844954e3909500c60eb1ca365ac9b5e7dc3d91122dbae46f402a5cef4e3b", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/runtime-planning/SKILL.md": "f0fe3d3f4053937767ae586559cc7fba89dacb6c896186a029fff220414e8d22", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/runtime-planning/references/memory-and-formats.md": "04edf10f4a965c512f6a262b1f8295f3c033d615849e49d21d3d94c78b549f2b", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/runtime-planning/references/troubleshooting.md": "5a1896f9fb01f6c8c9ef78d94f3c965d6d64e6b54223c6d7901eaed0548c5575", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/runtime-planning/scripts/parse_budget.py": "b3a3b0847280333c2bfbe5d03350dc866307d5290160ec64c7e3d462a34f6301", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/serving-cache/SKILL.md": "c16f911498e116ff9b31b5a6f0ddd3ae6517b51c0e9d079c337f36a5bcac1df8", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/serving-cache/references/state-and-api.md": "592603177473ac8e024b7dc19a3d3faedd7c1e0312137eab8bf9a62d2f567be6", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/serving-cache/references/troubleshooting.md": "4fda461042a3de43c109d7a3cec67542c7d61fe68412bf91606220f2e13e576e", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": "4c04d39812fb566e3ecb5807c170d5c32848eb2b93a1f0fedcf66ba9a9c92860", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": "13ab6d874c4e3135be80ef7c3cb6bfd28fb9121012e0536831ec3705ccb8c5eb", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/loader-storage-boundaries.md": "c1ec047aa37fca6b374f9bfd750e8f30d45af1d48fe728db23873c0b84ad94e8", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": "c493af8f6fd11ab7c9a37c9eec78cac31d57c04403a790035efbb4fff99f4669", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": "3aaf79b36192bf8ff3e8f95ac90f2e769aa065dfac1fe70368824f3591849b8d", + "src/inferrsi/data/skill_library/repositories/strata/LICENSE": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0", + "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed", + "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630", + "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b", + "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74", + "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0", + "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a", + "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397", + "src/inferrsi/data/skill_library/repositories/vllm-metal/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/repositories/vllm-metal/SKILL.md": "5f83b1b989e8dc239a6f34a6b3bd63872533ecc1404483a69dabfb1865648439", + "src/inferrsi/data/skill_library/repositories/vllm-metal/metadata.json": "63d51d1ace6b0ba503c28751a8ffbf9224038395f4146a94c775c628d8075470", + "src/inferrsi/data/skill_library/repositories/vllm-metal/references/repo-provenance.md": "970bda4d1081704aa82c94d6544a98bbf479430c4fc2c82a1a82f206043aa72a", + "src/inferrsi/data/skill_library/repositories/vllm-metal/references/repo-routing-metadata.json": "230a2b383bcf2855af4de50374c41f11b0b0841268ae45e424c1c97f679b477a", + "src/inferrsi/data/skill_library/repositories/vllm-metal/references/scope-and-notices.md": "aaef98eaea4cff56f39852e1d131163ebc01334e79034b1890f67a2660727d3e", + "src/inferrsi/data/skill_library/repositories/vllm-metal/references/source-identity.json": "9da529574e81f9f91d86c01153eb09a8b1e728dc5327c40364fff74085598c1f", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/SKILL.md": "dbfb06896615dc994f4a0a1e3789904acdf512ec316cbab6347165b764bc592f", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/references/cache-and-offload.md": "9829c14cf20f22575c44c1a0fcf18f63578443710e1a592c9fea709a43a3b888", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/references/helper-contract.md": "56a7494b58c47e07d1c85471ddc6075b15e34dc7a852b937002416dbd55b65ce", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/references/troubleshooting.md": "a4600c6d17a980b6442fc4c36f1c9ae148a2710c644126a9a898675b01f95366", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/scripts/screen_pool_budget.py": "dd615f7d3239f48510ce83416fb14fba8d5da2b9017627bdd312567be1bd7956", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/SKILL.md": "e2f5c5207b5c7665d3388d94fdace35e25f073a7833cbddaf14514a05f5ca5b9", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/references/configuration.md": "69949b93f4f319145e37d1b28487e37bd4ef21622f3169583d9cff46234a9c76", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/references/platform-and-loading.md": "7f11dc9e4bb9ae7bac25fcd6a2909cbea0eb2ae329e28ec9a7a2f4085c56e5cc", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/references/troubleshooting.md": "837cef82321fa050af3f01804fe337b0a0462255261fea5c21158ecfa3d0c325", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/scripts/check_environment_plan.py": "2603200fef87e3afad7820b84994f7e43e6041b45c17d0ed613d0d6229535bf2", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/SKILL.md": "37d9ae55ab4d0afd4ff8053c1955445e5a090188841ecc7941a5e22c5dd5eb39", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/references/dispatch-and-measurement.md": "360719d4253ad7e5a36823171d432aa8ca0f74d1ae565c7803af1990d99f2af9", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/references/record-schema.md": "f3cefb8b131613535dcacf30d130ffc2dea991acfb495adb4e1c7050046903d7", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/references/speculation-and-pipeline.md": "64faced7f31d489e9b0c4b9523a803ff1a76dca0f9315bdc6d0b83935bd338a4", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/references/troubleshooting.md": "a2e85696120ffc2af4249ac9c1e1b96320e02287e5099577c0474c025ab4f5f6", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/scripts/summarize_serving_records.py": "78e660a13a6a6941d006cb11fc2f401a34f489e54273860ae6a208dca6a6a3c2", + "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80", + "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220", + "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9", + "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/SKILL.md": "a468d75f25e77b96e6fb5376692f7a89c0464cc00d4df035f1dc6cc923f713cc", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/metadata.json": "db0b4593d7a6fbb898fdf11a1710b529abbade847ea57319843669b375b68fc8", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/references/NOTICE.md": "cca794656897ecdc37f983f7f23c4cd12306dbe3c9cd2f54fece6340753ae4cf", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/references/source-identity.json": "5e52db0a0bda4673b9aac46a3fc035b824780def72921fee6796edb2ed9fca65", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/references/source-notes.md": "0956f59cd9a1fac27da53228dd60c51a07b87639a43c09bfab296c4aeeceaab9", + "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a", + "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": "6f36cee8293b5484fd387a798cd450ad10cc2dc64aaded529e3a71ef4ecbe45f", + "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": "4c3f27459df239f9f4fc971c559d4c9ddfbfebd70507a8a93be8eae59a30b69d", + "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23", + "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": "d99fc28d48127d3bc627f2a0cab6f3f16158a77da49428838a08865adaea9905", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827", + "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1", + "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919", + "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941", + "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083", + "src/inferrsi/data/skill_library/skill-index.json": "b5a9f05bd8179644f3fea79d7eb04fe4a8baad9e8d68aa8d0f1db213cf633317", + "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226", + "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/metadata.json": "7a18f9dbb4894216802a44929b913990fc4bd8521cb60976d9435ee8f37e4265", + "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55", + "src/inferrsi/data/skill_library/workflows/claim-contract/metadata.json": "25c0137cf347f75d7ff8f8305d9ff5f055c6881ddf4b708f923c3176de16c153", + "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776", + "src/inferrsi/data/skill_library/workflows/competitor-scan/metadata.json": "74b387e5df7b40cce9fd1ceee7ea1a26c258a4ed091144f2b366ef971895a2e8", + "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56", + "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/metadata.json": "e97a665349f234d74baeaa20c6996bff990d0f45c40c0b7ac7283aa599cb951e", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": "073174963ca5c19c63d68843ccabdbf758f4c8b7ca61e3d7283affd6f88a753c", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": "08f76e12f6f8e8893f03bcdc8152a96c0c99442e7faecbc2a67a18c632fcb243", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/metadata.json": "caf3ef265df71b3cbec31b6c4affb6ebe9278fba0626b23bc124825d2761e49a", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/LICENSE-MIT.md": "1df1cf519b9e5d8acbbc72820299d7aa95afa1acbfc1f43a15109ad1d59975f3", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/NOTICE.md": "3c3c36a40e1d65f3a6b20ec3a97319cb753e857f1d120f302cd828e107120f6d", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/README.md": "c5802bb4d71fbf95e40a6a5282c0afcba594ab5b34d40bec97874928d8376a00", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/decision-cases.md": "c2c1d3d75c4fcb515dd36ff84d8dc940c25d54c5757a29a80d1af88eb8b66115", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/mechanisms.md": "4fa09099d362a8088fdb0e6a767f72aae6218b92fe5fd75b6fb78c2bb2a318fb", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/source-ledger.json": "60294951d702db924119300d89f4ed01d970a7036e214aa0d04c78f6842f92d9", + "src/inferrsi/evaluation.py": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e", + "src/inferrsi/evidence_paths.py": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa", + "src/inferrsi/execution.py": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92", + "src/inferrsi/harness/README.md": "0fd486efbf4752e9e93b118102244cb332ceea203dfa3628cb70fbf816b9a110", + "src/inferrsi/harness/__init__.py": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5", + "src/inferrsi/harness/evaluation.py": "20d0c52145600a5d6b1e513b61f2fffaa459b83db194876be40726bd4f343a59", + "src/inferrsi/harness/improvement.py": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8", + "src/inferrsi/harness/reporting.py": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3", + "src/inferrsi/harness/runner.py": "d290fc87710e77c8e6069e95bb0d347016a393707592d2eed98f07951f2d0865", + "src/inferrsi/harness/scenarios/__init__.py": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a", + "src/inferrsi/harness/scenarios/external.py": "f3f8d4fc516ebce81c26d82248e83f79b299eed0fc17ce73e2844995aea9c3f9", + "src/inferrsi/harness/scenarios/media.py": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a", + "src/inferrsi/harness/skill_eval_diagnostic.py": "e25cfee943052aaefbddf3f545fe6d56d95ab1edb9723cd9353feeabc954ea9a", + "src/inferrsi/harness/skill_eval_engineering.py": "2f18a1ab96acfc84319eba05e94c2a073fd0dca83e4f82ad98142ba2fecda039", + "src/inferrsi/harness/skill_evaluation.py": "ca141c3d1272591e24797d1ee31bed3aa37d9a009897c88c2ce0b087f3da11ac", + "src/inferrsi/infra/README.md": "8a3d41ba68cc655ef1788456e532125123902523f858b8932398995cc9127719", + "src/inferrsi/infra/__init__.py": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c", + "src/inferrsi/infra/artifacts.py": "bcf3adf325e34e10a4ecb4baf0af855ddca3d77eff76de5b66e493baaa3d7eaf", + "src/inferrsi/infra/campaign.py": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209", + "src/inferrsi/infra/commands.py": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1", + "src/inferrsi/infra/environment.py": "23a6716089c542617761b5a3fc67dde0af9e483212fc003f8577757fbc4ed2db", + "src/inferrsi/infra/executors/__init__.py": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a", + "src/inferrsi/infra/executors/local.py": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77", + "src/inferrsi/infra/http.py": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2", + "src/inferrsi/infra/inspection.py": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd", + "src/inferrsi/infra/platforms/__init__.py": "dac68eb6edbff995f8afe477396d81efd5db93128978ceedaea1babf0dd3db08", + "src/inferrsi/infra/platforms/cuda.py": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f", + "src/inferrsi/infra/platforms/mlx.py": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3", + "src/inferrsi/infra/processes.py": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189", + "src/inferrsi/infra/targets/__init__.py": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3", + "src/inferrsi/infra/targets/git.py": "abb93f9332db336298102e28c06b9a7553597d4d03b957358602f60d61b1aa65", + "src/inferrsi/inspection.py": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3", + "src/inferrsi/knowledge_operations.py": "c1a5f0756cc1cdad66d7b766f68405b622d85d65311546a952ab77b66f0bde44", + "src/inferrsi/knowledge_store.py": "70dfca059b0a682579206d4376d191667934b3ca869ae847098c4c6964d86abe", + "src/inferrsi/model/README.md": "a8a4fefbc943c97b2fac6c4a05e751c90b1e79d46de8fb59f500e92f8ddd5e34", + "src/inferrsi/model/__init__.py": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f", + "src/inferrsi/model/anthropic_agent.py": "746db742467fee153427cc9d7e8a01b640bc13e5b9cd431e935f596a8e19b072", + "src/inferrsi/model/protocols/__init__.py": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47", + "src/inferrsi/model/protocols/image.py": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f", + "src/inferrsi/model/protocols/metrics.py": "5b32e296a746b6de30cb0c89f23f730296b0087884c1acc2b6127c1295d4db7a", + "src/inferrsi/model/request_guard.py": "c2f8a2f822f32785327a2138ee7b0609896b180819d9d1cbe851534aa6198628", + "src/inferrsi/model/validators/__init__.py": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b", + "src/inferrsi/model/validators/flare.py": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f", + "src/inferrsi/model/validators/image.py": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a", + "src/inferrsi/reporting.py": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d", + "src/inferrsi/research.py": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04", + "src/inferrsi/runspec.py": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709", + "src/inferrsi/skill_catalog.py": "dbbe107a105e13b0c288f48ad80f9b05e108606af379d243dfab85cdfc99cfb5", + "src/inferrsi/summary_chart.py": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc" + } +} diff --git a/evidence/indexer-255-covered-cue-20261030/results/consolidated-results.json b/evidence/indexer-255-covered-cue-20261030/results/consolidated-results.json new file mode 100644 index 000000000..1fb68feba --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/results/consolidated-results.json @@ -0,0 +1,74 @@ +[ + { + "study": "Efficiency v1", + "baseline": "ab03f3ffb8f2e5bc02c304828c5c41890eef673e", + "candidate": "6da28494891b21cdc89f632d84d763bc997b7d15", + "protocol": "evidence/indexer-255-20261029/plan/protocol.json", + "report": "evidence/indexer-255-20261029/results/report.md", + "counts": "20 paired AB/BA groups ×10 paths; primary4 queries; accuracy8DEV+4heldout;6contracts", + "result": "FAIL: primary median reduction58.6105% (95%CI52.5395–61.3458%); CLIsearchp95+52.7193% exceeds20%guard", + "publication": "Published in #256, merge3731b731; runtime not promoted", + "protocol_sha256": "e4964329a20e6752b8149ab270733c0bbb258d0f948bdca9e86da6fe8f028956", + "report_sha256": "38186030ae6cf8a94d202c4c1d6ae41bad8b19d9060f63dd350ae8c66896cb2a" + }, + { + "study": "Efficiency confirmation v2 (not accuracy v2)", + "baseline": "ab03f3ffb8f2e5bc02c304828c5c41890eef673e", + "candidate": "6da28494891b21cdc89f632d84d763bc997b7d15", + "protocol": "evidence/indexer-255-20261029/plan/confirmation-v2.json", + "report": "evidence/indexer-255-20261029/results/report.md", + "counts": "120pairedCLIsearch samples;40separatebaseline/baselinecontrolpairs", + "result": "PointgatesPASS only: p50+9.9734%, p95−3.2103%;95%CIp50+1.2014…+20.9187%,p95−22.2255…+50.8618%; v1FAILunchanged, no promotion", + "publication": "Published in #256, same historical campaign; no pooled/replacement PASS", + "protocol_sha256": "40de4eefba8712d3d07f2e48049b0899c831f94bfcfa252c1f8841525a337fa4", + "report_sha256": "38186030ae6cf8a94d202c4c1d6ae41bad8b19d9060f63dd350ae8c66896cb2a" + }, + { + "study": "Accuracy whole-token v2", + "baseline": "1147c5bb41724826b64ad13cebc8a4dce5cc278b", + "candidate": "55a458328195335586eb6c6c1301ba34a8218e63", + "protocol": "evidence/indexer-255-token-v2-20261030/plan/accepted-protocol.json", + "report": "evidence/indexer-255-token-v2-20261030/results/report.md", + "counts": "6DEV+2nongatingdomaincases;8synthetic;6sealedheldoutNOTRUN", + "result": "FAIL: Hit4/6→5/6;MRR7/12(.583333)→13/24(.541667);complete4→5;38contexts/arm fidelityPASS", + "publication": "Published in #259, mergedabd1849; candidatepatch/bundle only", + "protocol_sha256": "6e72cf159003a1e85b0b4cdc42beee895494da294162e19adecfe963613de517", + "report_sha256": "4339cc0733ef8bed0a402f8b8008bfcaf3bdf5f2e1036ebd9efe9abeb0f6e347" + }, + { + "study": "Accuracy structured-string v3", + "baseline": "b3549ce1aea1556f76de1d2e14a4240474947b6b", + "candidate": "96788d26b57240e53c26fe5446b2062a62ee6d0e", + "protocol": "evidence/indexer-255-token-v3-20261030/plan/protocol.md", + "report": "evidence/indexer-255-token-v3-20261030/results/report.md", + "counts": "6reusedtrainingDEV+2domaincases;18synthetic;6sealedheldoutNOTRUN", + "result": "FAIL: Hit4/6→5/6;MRR7/12→8/15(.533333);complete4→5;51testsPASS;38contexts/arm fidelityPASS", + "publication": "Unpublished local complete evidence + candidatepatch/bundle; no runtimepromotion", + "protocol_sha256": "2adf6a226b4c3763e52dec159aaae9ec4ea0824f2b0b0bd37a92799ef06bf94d", + "report_sha256": "b58be2e52508d425ab5e0b34bf2bdb107bb32d8caff8749f1d194a16b520be8a" + }, + { + "study": "Blind-cue preparation", + "baseline": "b3549ce1aea1556f76de1d2e14a4240474947b6b", + "candidate": "None; production unchanged, no retrieval", + "protocol": "evidence/indexer-255-blind-cue-20261030/plan/protocol-draft.md", + "report": "evidence/indexer-255-blind-cue-20261030/results/source-label-review.json", + "counts": "6authoredraw/cuepairs;3source-supported <4minimum;0pairedretrievalruns", + "result": "PREPARATION INCONCLUSIVE: intent reviewpassed but coveragegatefailed; no replacements or effect claim", + "publication": "Unpublished preparation receipts and inputs; no retrievaldata to invent", + "protocol_sha256": "0dbe4843e50fc064934e98cad369024ae239b36c69772e2071d7e96c4eecbf17", + "report_sha256": "3ca52b5576fca03c7e536d6c19b4ff2ff2b0c88c6c8a63dfcdfdf2d5de5f586f" + }, + { + "study": "Source-conditioned covered-cue", + "baseline": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "candidate": "None; identicalproductionruntime, cueinputonly", + "protocol": "evidence/indexer-255-covered-cue-20261030/plan/protocol-draft.md", + "report": "evidence/indexer-255-covered-cue-20261030/results/report.md", + "counts": "6coveredsyntheticpairs/12captures;2byteidenticalcontrols;4mostlyframingedits", + "result": "Neutral FAIL: Hit6/6→6/6;MRR29/36→29/36;complete6→6;strictMRRgainfalse;29contexts/arm fidelityPASS", + "publication": "Unpublished frozenprep,runner,rawgzip,metrics,independentaudit; productionunchanged", + "protocol_sha256": "4512e1ea3d56235b207e4926c0ed84e8eb3022a461830c919ad29810ee27b026", + "report_sha256": "7af98da8eba7319c5d853b3c7932ab54dd434df50edb221c5de15aa98e7e7bd8" + } +] diff --git a/evidence/indexer-255-covered-cue-20261030/results/consolidated-results.md b/evidence/indexer-255-covered-cue-20261030/results/consolidated-results.md new file mode 100644 index 000000000..14a58cbf6 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/results/consolidated-results.md @@ -0,0 +1,16 @@ +# #255 consolidated results + +Evidence delivery is not accepted efficiency/accuracy gain. Keep #255 open. This table consolidates existing artifacts only; no new experiment was run for packaging. Historical outcomes and distinct baselines remain intact. + +| Study | Samples | Result | Evidence | +|---|---|---|---| +| Efficiency v1 | 20 paired AB/BA groups ×10 paths; primary4 queries; accuracy8DEV+4heldout;6contracts | FAIL: primary median reduction58.6105% (95%CI52.5395–61.3458%); CLIsearchp95+52.7193% exceeds20%guard | [Report](../../../evidence/indexer-255-20261029/results/report.md) | +| Efficiency confirmation v2 (not accuracy v2) | 120pairedCLIsearch samples;40separatebaseline/baselinecontrolpairs | PointgatesPASS only: p50+9.9734%, p95−3.2103%;95%CIp50+1.2014…+20.9187%,p95−22.2255…+50.8618%; v1FAILunchanged, no promotion | [Report](../../../evidence/indexer-255-20261029/results/report.md) | +| Accuracy whole-token v2 | 6DEV+2nongatingdomaincases;8synthetic;6sealedheldoutNOTRUN | FAIL: Hit4/6→5/6;MRR7/12(.583333)→13/24(.541667);complete4→5;38contexts/arm fidelityPASS | [Report](../../../evidence/indexer-255-token-v2-20261030/results/report.md) | +| Accuracy structured-string v3 | 6reusedtrainingDEV+2domaincases;18synthetic;6sealedheldoutNOTRUN | FAIL: Hit4/6→5/6;MRR7/12→8/15(.533333);complete4→5;51testsPASS;38contexts/arm fidelityPASS | [Report](../../../evidence/indexer-255-token-v3-20261030/results/report.md) | +| Blind-cue preparation | 6authoredraw/cuepairs;3source-supported <4minimum;0pairedretrievalruns | PREPARATION INCONCLUSIVE: intent reviewpassed but coveragegatefailed; no replacements or effect claim | [Report](../../../evidence/indexer-255-blind-cue-20261030/results/source-label-review.json) | +| Source-conditioned covered-cue | 6coveredsyntheticpairs/12captures;2byteidenticalcontrols;4mostlyframingedits | Neutral FAIL: Hit6/6→6/6;MRR29/36→29/36;complete6→6;strictMRRgainfalse;29contexts/arm fidelityPASS | [Report](../../../evidence/indexer-255-covered-cue-20261030/results/report.md) | + +Exact baselines, candidate revisions, protocol/report repository paths and SHA256 are in [consolidated-results.json](consolidated-results.json). Timing confirmation v2 is distinct from accuracy v2: its point-gate pass and broad uncertainty never replace timingv1FAIL. Accuracyv3 DEV is known training diagnostics, not independent validation. The newer sealed six-case heldout was never evaluated or exposed to the implementer; its authors/source reviewers necessarily accessed it for preparation. It differs from the historical four-case heldout used in the initial efficiency assessment. No sealed content is included here. + +Published campaigns #256/#259 are linked, not modified. This follow-up archives v3FAIL, blind-cue preparationINCONCLUSIVE and covered-cue neutralFAIL together. No production source change, runtime promotion, new algorithm, cost or issue-closure claim. Partial positives do not establish complete Recall/precision, population accuracy, speed nonregression, or downstream-agent benefit. Any further hypothesis is a separate prospective decision. diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-01-cue.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-01-cue.json.gz new file mode 100644 index 000000000..f5bbc89cc Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-01-cue.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-01-raw.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-01-raw.json.gz new file mode 100644 index 000000000..c59b1cb47 Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-01-raw.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-02-cue.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-02-cue.json.gz new file mode 100644 index 000000000..49885f8bc Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-02-cue.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-02-raw.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-02-raw.json.gz new file mode 100644 index 000000000..49885f8bc Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-02-raw.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-03-cue.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-03-cue.json.gz new file mode 100644 index 000000000..512c097f1 Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-03-cue.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-03-raw.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-03-raw.json.gz new file mode 100644 index 000000000..512c097f1 Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-03-raw.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-04-cue.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-04-cue.json.gz new file mode 100644 index 000000000..15104000e Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-04-cue.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-04-raw.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-04-raw.json.gz new file mode 100644 index 000000000..cc7034c23 Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-04-raw.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-05-cue.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-05-cue.json.gz new file mode 100644 index 000000000..8ad75c3b5 Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-05-cue.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-05-raw.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-05-raw.json.gz new file mode 100644 index 000000000..323478d33 Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-05-raw.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-06-cue.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-06-cue.json.gz new file mode 100644 index 000000000..abe33853a Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-06-cue.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/covered255-06-raw.json.gz b/evidence/indexer-255-covered-cue-20261030/results/covered255-06-raw.json.gz new file mode 100644 index 000000000..cf89b2ec8 Binary files /dev/null and b/evidence/indexer-255-covered-cue-20261030/results/covered255-06-raw.json.gz differ diff --git a/evidence/indexer-255-covered-cue-20261030/results/default-captures.json b/evidence/indexer-255-covered-cue-20261030/results/default-captures.json new file mode 100644 index 000000000..9deff6f21 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/results/default-captures.json @@ -0,0 +1,108 @@ +{ + "raw": [ + { + "options": {}, + "sha256": "3ce21430de7651836db033a90dc8f3d930a70fffcd8177f98a72ff21b3046788", + "bytes": 518449 + }, + { + "options": { + "relevance": "cgroupv2", + "context_bytes": 262144 + }, + "sha256": "28007091e8ab7931bbc4e56ada20aa63a93d6b41d39f33950db8789e11118037", + "bytes": 167350 + }, + { + "options": { + "query": "stat.loc", + "relevance": "cgroupv2" + }, + "sha256": "fb72788b1bc00b34ebb34281cc5b25848951540c344294798d2941e3d73476bd", + "bytes": 156919 + }, + { + "options": { + "relevance": "gpu", + "backend": "ascend" + }, + "sha256": "a7aa9faf34fb38a84d5a7b2ce125dee9cf31a6ce891e3fcd608e89c102cda48f", + "bytes": 1583960 + }, + { + "options": { + "ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "relevance": "ram", + "context_bytes": 262144 + }, + "sha256": "383950753b8454c5916018cfd72e37b9c66c2abc4ddef9bfb8facfc306627404", + "bytes": 167346 + }, + { + "options": { + "ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "relevance": "ram", + "context_bytes": 1 + }, + "sha256": "eca670ae06bbf9b45b73831bd706dd023e9bcb741e2cd2b7e42f78b45bca72b6", + "bytes": 157247 + } + ], + "cue": [ + { + "options": {}, + "sha256": "3ce21430de7651836db033a90dc8f3d930a70fffcd8177f98a72ff21b3046788", + "bytes": 518449 + }, + { + "options": { + "relevance": "cgroupv2", + "context_bytes": 262144 + }, + "sha256": "28007091e8ab7931bbc4e56ada20aa63a93d6b41d39f33950db8789e11118037", + "bytes": 167350 + }, + { + "options": { + "query": "stat.loc", + "relevance": "cgroupv2" + }, + "sha256": "fb72788b1bc00b34ebb34281cc5b25848951540c344294798d2941e3d73476bd", + "bytes": 156919 + }, + { + "options": { + "relevance": "gpu", + "backend": "ascend" + }, + "sha256": "a7aa9faf34fb38a84d5a7b2ce125dee9cf31a6ce891e3fcd608e89c102cda48f", + "bytes": 1583960 + }, + { + "options": { + "ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "relevance": "ram", + "context_bytes": 262144 + }, + "sha256": "383950753b8454c5916018cfd72e37b9c66c2abc4ddef9bfb8facfc306627404", + "bytes": 167346 + }, + { + "options": { + "ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "relevance": "ram", + "context_bytes": 1 + }, + "sha256": "eca670ae06bbf9b45b73831bd706dd023e9bcb741e2cd2b7e42f78b45bca72b6", + "bytes": 157247 + } + ] +} diff --git a/evidence/indexer-255-covered-cue-20261030/results/independent-review.json b/evidence/indexer-255-covered-cue-20261030/results/independent-review.json new file mode 100644 index 000000000..92c081867 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/results/independent-review.json @@ -0,0 +1,85 @@ +{ + "reviewed_at_utc": "2026-10-30T03:49:44.919304+00:00", + "baseline": "6e6ee245d9e92ee9365976f2e4e262304999f147", + "scope": "Independent retained-output audit only, no collect/retrieval rerun. Pinned reader.read and indexed_skill used to verify saved resource sets.", + "metrics": { + "raw": { + "ranks": [ + 2, + 3, + 1, + 1, + 1, + 1 + ], + "complete_relevant": 6, + "contexts": 29, + "hit": 6, + "mrr_exact": "29/36" + }, + "cue": { + "ranks": [ + 2, + 3, + 1, + 1, + 1, + 1 + ], + "complete_relevant": 6, + "contexts": 29, + "hit": 6, + "mrr_exact": "29/36" + } + }, + "checks": [ + { + "id": "covered255-01", + "identity_control": false, + "raw_sha256": "1a4cf2640a0ce5617b9a3b1f75b4335d1c0fe9f7186b055ce634dfeedd1a7446", + "cue_sha256": "0174afc59d1af3d0a77f5c99eaa93ed77fe3ed6865a6c67e96524e70d2217052" + }, + { + "id": "covered255-02", + "identity_control": true, + "raw_sha256": "817f8961b8a310682134a07d7bd688c894f72291554a19f7b901573e872febb0", + "cue_sha256": "817f8961b8a310682134a07d7bd688c894f72291554a19f7b901573e872febb0" + }, + { + "id": "covered255-03", + "identity_control": true, + "raw_sha256": "d9c629f578fce712d39816e8be8775ac060f4fc8f08821bfff01d3952d1bbbaf", + "cue_sha256": "d9c629f578fce712d39816e8be8775ac060f4fc8f08821bfff01d3952d1bbbaf" + }, + { + "id": "covered255-04", + "identity_control": false, + "raw_sha256": "18eb7cbaee69d9795498a83aed857ab1dc27af8d779242622dde650231a465fa", + "cue_sha256": "a8d411c2b76f0ad14a8663d25e2d1be8d4efb20285074c9ea6c482c6a299d6e9" + }, + { + "id": "covered255-05", + "identity_control": false, + "raw_sha256": "1d3a8d8b3098b7cee4d7a4553e91420adb041dbceb05c975884451da50371c8c", + "cue_sha256": "f744bb7430d2c79786ec30b912af2a84f04c95613b938be82b56fde007c924d9" + }, + { + "id": "covered255-06", + "identity_control": false, + "raw_sha256": "161f56d917c53d40a73de6b9b797622d1daf6f128d9add0eab9ae0d90ab642d1", + "cue_sha256": "4a5366c58b109a22bd160c74664d8cad12aac1502cc13a022f736a0e0d07072c" + } + ], + "source_input_runner_hashes_verified": true, + "identity_controls_exact_parity": true, + "heldout": "UNUSED", + "verdict": "FAIL_STRICT_MRR_IMPROVEMENT_GATE", + "limits": [ + "Neutral labelled ranking outcome, not pipeline failure or proof all outputs unchanged.", + "Source-conditioned synthetic cases and four politeness-only removals; no population/automatic-normalizer inference.", + "Previous blind-input preparationINCONCLUSIVE remains separate." + ], + "default_capture_parity": true, + "default_capture_count": 6, + "default_verification_limit": "Equality of retained capture receipts; no rerun of default calls." +} diff --git a/evidence/indexer-255-covered-cue-20261030/results/partial-results.json b/evidence/indexer-255-covered-cue-20261030/results/partial-results.json new file mode 100644 index 000000000..bd79fa137 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/results/partial-results.json @@ -0,0 +1,283 @@ +{ + "rows": [ + { + "id": "covered255-01", + "identity_control": false, + "arms": { + "raw": { + "rank": 2, + "ids": [ + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.cgroup_resource_controls", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.common.skills.video_diffusion_optimization" + ], + "complete_relevant": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "audited_contexts": 5, + "output_bytes": 1083382, + "context_bytes": 256659, + "context_omissions": [], + "sha256": "1a4cf2640a0ce5617b9a3b1f75b4335d1c0fe9f7186b055ce634dfeedd1a7446", + "artifact": "covered255-01-raw.json.gz" + }, + "cue": { + "rank": 2, + "ids": [ + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.cgroup_resource_controls", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.common.skills.dense_performance_evidence" + ], + "complete_relevant": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "audited_contexts": 5, + "output_bytes": 1081318, + "context_bytes": 256138, + "context_omissions": [], + "sha256": "0174afc59d1af3d0a77f5c99eaa93ed77fe3ed6865a6c67e96524e70d2217052", + "artifact": "covered255-01-cue.json.gz" + } + } + }, + { + "id": "covered255-02", + "identity_control": true, + "arms": { + "raw": { + "rank": 3, + "ids": [ + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.ascend.skills.ascend_performance_analysis", + "inferrsi.ascend.skills.ascend_device_health", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.dense_performance_evidence" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.ascend_device_health" + ], + "audited_contexts": 5, + "output_bytes": 1057891, + "context_bytes": 234561, + "context_omissions": [], + "sha256": "817f8961b8a310682134a07d7bd688c894f72291554a19f7b901573e872febb0", + "artifact": "covered255-02-raw.json.gz" + }, + "cue": { + "rank": 3, + "ids": [ + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.ascend.skills.ascend_performance_analysis", + "inferrsi.ascend.skills.ascend_device_health", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.dense_performance_evidence" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.ascend_device_health" + ], + "audited_contexts": 5, + "output_bytes": 1057891, + "context_bytes": 234561, + "context_omissions": [], + "sha256": "817f8961b8a310682134a07d7bd688c894f72291554a19f7b901573e872febb0", + "artifact": "covered255-02-cue.json.gz" + } + } + }, + { + "id": "covered255-03", + "identity_control": true, + "arms": { + "raw": { + "rank": 1, + "ids": [ + "inferrsi.common.skills.os_process_placement", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.bound_windowed_attention_replay" + ], + "complete_relevant": [ + "inferrsi.common.skills.os_process_placement" + ], + "audited_contexts": 5, + "output_bytes": 961583, + "context_bytes": 146747, + "context_omissions": [], + "sha256": "d9c629f578fce712d39816e8be8775ac060f4fc8f08821bfff01d3952d1bbbaf", + "artifact": "covered255-03-raw.json.gz" + }, + "cue": { + "rank": 1, + "ids": [ + "inferrsi.common.skills.os_process_placement", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.bound_windowed_attention_replay" + ], + "complete_relevant": [ + "inferrsi.common.skills.os_process_placement" + ], + "audited_contexts": 5, + "output_bytes": 961583, + "context_bytes": 146747, + "context_omissions": [], + "sha256": "d9c629f578fce712d39816e8be8775ac060f4fc8f08821bfff01d3952d1bbbaf", + "artifact": "covered255-03-cue.json.gz" + } + } + }, + { + "id": "covered255-04", + "identity_control": false, + "arms": { + "raw": { + "rank": 1, + "ids": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.ascend.skills.ascend_graph_mode", + "inferrsi.ascend.skills.cann_attribution_overlap" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay" + ], + "audited_contexts": 5, + "output_bytes": 961468, + "context_bytes": 144355, + "context_omissions": [], + "sha256": "18eb7cbaee69d9795498a83aed857ab1dc27af8d779242622dde650231a465fa", + "artifact": "covered255-04-raw.json.gz" + }, + "cue": { + "rank": 1, + "ids": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.ascend.skills.ascend_graph_mode", + "inferrsi.ascend.skills.cann_attribution_overlap" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay" + ], + "audited_contexts": 5, + "output_bytes": 961442, + "context_bytes": 144355, + "context_omissions": [], + "sha256": "a8d411c2b76f0ad14a8663d25e2d1be8d4efb20285074c9ea6c482c6a299d6e9", + "artifact": "covered255-04-cue.json.gz" + } + } + }, + { + "id": "covered255-05", + "identity_control": false, + "arms": { + "raw": { + "rank": 1, + "ids": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle", + "inferrsi.common.skills.host_memory_management", + "inferrsi.common.skills.bound_windowed_attention_replay" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle" + ], + "audited_contexts": 5, + "output_bytes": 1006115, + "context_bytes": 183756, + "context_omissions": [], + "sha256": "1d3a8d8b3098b7cee4d7a4553e91420adb041dbceb05c975884451da50371c8c", + "artifact": "covered255-05-raw.json.gz" + }, + "cue": { + "rank": 1, + "ids": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle", + "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.host_memory_management", + "inferrsi.common.skills.bound_windowed_attention_replay" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle" + ], + "audited_contexts": 5, + "output_bytes": 1006078, + "context_bytes": 183756, + "context_omissions": [], + "sha256": "f744bb7430d2c79786ec30b912af2a84f04c95613b938be82b56fde007c924d9", + "artifact": "covered255-05-cue.json.gz" + } + } + }, + { + "id": "covered255-06", + "identity_control": false, + "arms": { + "raw": { + "rank": 1, + "ids": [ + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.common.skills.checkpoint_recovery_consistency", + "inferrsi.common.skills.performance_ceiling_analysis" + ], + "complete_relevant": [ + "inferrsi.common.skills.weight_loading_optimization" + ], + "audited_contexts": 4, + "output_bytes": 961055, + "context_bytes": 143629, + "context_omissions": [ + { + "ref": "package:inferrsi.common.skills.performance_ceiling_analysis", + "reason": "context_limit", + "required_bytes": 170782, + "available_bytes": 118515 + } + ], + "sha256": "161f56d917c53d40a73de6b9b797622d1daf6f128d9add0eab9ae0d90ab642d1", + "artifact": "covered255-06-raw.json.gz" + }, + "cue": { + "rank": 1, + "ids": [ + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.ascend.skills.cann_attribution_overlap" + ], + "complete_relevant": [ + "inferrsi.common.skills.weight_loading_optimization" + ], + "audited_contexts": 4, + "output_bytes": 945736, + "context_bytes": 129803, + "context_omissions": [ + { + "ref": "package:inferrsi.common.skills.performance_ceiling_analysis", + "reason": "context_limit", + "required_bytes": 170782, + "available_bytes": 141313 + } + ], + "sha256": "4a5366c58b109a22bd160c74664d8cad12aac1502cc13a022f736a0e0d07072c", + "artifact": "covered255-06-cue.json.gz" + } + } + } + ], + "errors": [] +} diff --git a/evidence/indexer-255-covered-cue-20261030/results/report.md b/evidence/indexer-255-covered-cue-20261030/results/report.md new file mode 100644 index 000000000..334c5b060 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/results/report.md @@ -0,0 +1,39 @@ +# #255 source-conditioned cue feasibility — FAIL + +## Decision +The one prospectively frozen paired run did not improve retrieval ranks: raw task and verbatim cue each achieve partial-label Hit@5 **6/6**, MRR@5 **29/36** (approximately0.805556), and complete relevant delivery **6/6**. The strict MRR improvement gate fails. No further wording changes or retries were made; no production change, promotion, or #255 closure follows. + +All other evaluated gates pass: no labelled hits lost; identical-input controls produce identical serialized outputs; six unchanged-interface captures match; full required-resource integrity and budget accounting pass; frozen source remained unchanged; execution recorded no errors. + +## What was actually tested +Six deliberately source-covered, synthetic tasks authored from reviewed source capsules. This is source-conditioned feasibility, not organic demand or unbiased general retrieval accuracy. Four cue edits mostly remove polite introductory framing and join retained spans with semicolons. Two controls are byte-identical. This narrow intervention does not establish broad narrative normalization, automatic rewriting, cheap human preparation, downstream agent benefit, or coverage of arbitrary repository tasks. + +The previous blind-task preparation-INCONCLUSIVE (only three supported pairs below its four-case minimum) and prior performance/v2/v3 FAIL results remain separate and unchanged. The old sealed heldout was not accessed or evaluated. + +## Frozen identity / execution +- Runtime/corpus baseline: `6e6ee245d9e92ee9365976f2e4e262304999f147`, production weighted-lexical-v1, unchanged in both arms. +- Accepted protocol: `4512e1ea3d56235b207e4926c0ed84e8eb3022a461830c919ad29810ee27b026`. +- Inputs: `24e822eaadcd4598435dac0e7cf77cc1dd73d8a6c50bc0f7c5982253b4a94fcc`. +- Input freeze: `305246e2c49ea47bea9dbe9dc412e5a8920acec735f01f445514938b8cf7b9c3`. +- Frozen runner: `566c9e2d3a63eb3d9b7e8f4b1e41ce1c5121e316c0ffa50ffc01fd0aca40c1af`; runner-freeze.json predates all retrieval. +- Exactly one capture per arm/case through production Collect with query empty, no filters, limit5, context budget262144. No new scorer, source edits, installations, models/services, hardware or timing campaign. + +## Per-case table +| Case | Raw rank | Cue rank | Identity control | Raw/cue chars | Context bytes raw/cue | Output bytes raw/cue | +|---|---:|---:|---|---:|---:|---:| +| covered255-01 | 2 | 2 | False | 535/497 | 256659/256138 | 1083382/1081318 | +| covered255-02 | 3 | 3 | True | 299/299 | 234561/234561 | 1057891/1057891 | +| covered255-03 | 1 | 1 | True | 275/275 | 146747/146747 | 961583/961583 | +| covered255-04 | 1 | 1 | False | 558/532 | 144355/144355 | 961468/961442 | +| covered255-05 | 1 | 1 | False | 592/555 | 183756/183756 | 1006115/1006078 | +| covered255-06 | 1 | 1 | False | 521/487 | 143629/129803 | 961055/945736 | + +Ranks are over partial positive IDs, not exhaustive precision/recall labels. Complete relevant delivery is audited separately from rank. Output byte differences may include selection cue metadata and do not imply rank or reasoning improvement. Full raw outputs preserve item order, resources and omissions. + +## Independent supplemental audit +`independent-review.json` verifies all12 saved outputs, exact ranks/fractional arithmetic,29 emitted contexts per arm with full required sets/bytes/hash/revision/budget, unchanged frozen identities and control/default parity. No retrieval or heldout was rerun. Some unlabelled top-five items change, so this is neutral on labelled rank metrics, not a claim that all retrieval outputs are identical. + +## Artifacts / replay +`results.json` contains exact fractional metrics, gate booleans, per-case summaries and errors. `partial-results.json` preserves incremental captures; final state is complete. Twelve lossless deterministic gzip JSON files retain complete outputs; their decompressed SHA256 and lengths appear in results.json. `default-captures.json` preserves the six unchanged-interface scenarios for each arm. `run.log` records terminal verdict. Protocol/input/source manifests remain in the separately frozen preparation directory. + +For replay only under an authorized independent study, use existing Python/PyYAML: create a new output directory, then `python evaluate.py BASELINE_ROOT PREP_ROOT OUTPUT_DIR`. Do not overwrite frozen outputs or reinterpret a replay as a new successful experiment. No speed/nonregression claim follows from this untimed CPU-only trial. diff --git a/evidence/indexer-255-covered-cue-20261030/results/results.json b/evidence/indexer-255-covered-cue-20261030/results/results.json new file mode 100644 index 000000000..bc488f135 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/results/results.json @@ -0,0 +1,306 @@ +{ + "verdict": "FAIL", + "metrics": { + "raw": { + "hit": 6, + "mrr_exact": "29/36", + "complete_relevant": 6 + }, + "cue": { + "hit": 6, + "mrr_exact": "29/36", + "complete_relevant": 6 + } + }, + "gates": { + "no_hit_lost": true, + "hit_nondecreasing": true, + "mrr_strictly_greater": false, + "complete_nondecreasing": true, + "identity_control_parity": true, + "default_parity": true, + "resource_audits": true + }, + "source_unchanged": true, + "rows": [ + { + "id": "covered255-01", + "identity_control": false, + "arms": { + "raw": { + "rank": 2, + "ids": [ + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.cgroup_resource_controls", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.common.skills.video_diffusion_optimization" + ], + "complete_relevant": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "audited_contexts": 5, + "output_bytes": 1083382, + "context_bytes": 256659, + "context_omissions": [], + "sha256": "1a4cf2640a0ce5617b9a3b1f75b4335d1c0fe9f7186b055ce634dfeedd1a7446", + "artifact": "covered255-01-raw.json.gz" + }, + "cue": { + "rank": 2, + "ids": [ + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.cgroup_resource_controls", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.common.skills.dense_performance_evidence" + ], + "complete_relevant": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "audited_contexts": 5, + "output_bytes": 1081318, + "context_bytes": 256138, + "context_omissions": [], + "sha256": "0174afc59d1af3d0a77f5c99eaa93ed77fe3ed6865a6c67e96524e70d2217052", + "artifact": "covered255-01-cue.json.gz" + } + } + }, + { + "id": "covered255-02", + "identity_control": true, + "arms": { + "raw": { + "rank": 3, + "ids": [ + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.ascend.skills.ascend_performance_analysis", + "inferrsi.ascend.skills.ascend_device_health", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.dense_performance_evidence" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.ascend_device_health" + ], + "audited_contexts": 5, + "output_bytes": 1057891, + "context_bytes": 234561, + "context_omissions": [], + "sha256": "817f8961b8a310682134a07d7bd688c894f72291554a19f7b901573e872febb0", + "artifact": "covered255-02-raw.json.gz" + }, + "cue": { + "rank": 3, + "ids": [ + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.ascend.skills.ascend_performance_analysis", + "inferrsi.ascend.skills.ascend_device_health", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.dense_performance_evidence" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.ascend_device_health" + ], + "audited_contexts": 5, + "output_bytes": 1057891, + "context_bytes": 234561, + "context_omissions": [], + "sha256": "817f8961b8a310682134a07d7bd688c894f72291554a19f7b901573e872febb0", + "artifact": "covered255-02-cue.json.gz" + } + } + }, + { + "id": "covered255-03", + "identity_control": true, + "arms": { + "raw": { + "rank": 1, + "ids": [ + "inferrsi.common.skills.os_process_placement", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.bound_windowed_attention_replay" + ], + "complete_relevant": [ + "inferrsi.common.skills.os_process_placement" + ], + "audited_contexts": 5, + "output_bytes": 961583, + "context_bytes": 146747, + "context_omissions": [], + "sha256": "d9c629f578fce712d39816e8be8775ac060f4fc8f08821bfff01d3952d1bbbaf", + "artifact": "covered255-03-raw.json.gz" + }, + "cue": { + "rank": 1, + "ids": [ + "inferrsi.common.skills.os_process_placement", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.bound_windowed_attention_replay" + ], + "complete_relevant": [ + "inferrsi.common.skills.os_process_placement" + ], + "audited_contexts": 5, + "output_bytes": 961583, + "context_bytes": 146747, + "context_omissions": [], + "sha256": "d9c629f578fce712d39816e8be8775ac060f4fc8f08821bfff01d3952d1bbbaf", + "artifact": "covered255-03-cue.json.gz" + } + } + }, + { + "id": "covered255-04", + "identity_control": false, + "arms": { + "raw": { + "rank": 1, + "ids": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.ascend.skills.ascend_graph_mode", + "inferrsi.ascend.skills.cann_attribution_overlap" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay" + ], + "audited_contexts": 5, + "output_bytes": 961468, + "context_bytes": 144355, + "context_omissions": [], + "sha256": "18eb7cbaee69d9795498a83aed857ab1dc27af8d779242622dde650231a465fa", + "artifact": "covered255-04-raw.json.gz" + }, + "cue": { + "rank": 1, + "ids": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.ascend.skills.ascend_graph_mode", + "inferrsi.ascend.skills.cann_attribution_overlap" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.ascend_aclgraph_capture_replay" + ], + "audited_contexts": 5, + "output_bytes": 961442, + "context_bytes": 144355, + "context_omissions": [], + "sha256": "a8d411c2b76f0ad14a8663d25e2d1be8d4efb20285074c9ea6c482c6a299d6e9", + "artifact": "covered255-04-cue.json.gz" + } + } + }, + { + "id": "covered255-05", + "identity_control": false, + "arms": { + "raw": { + "rank": 1, + "ids": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle", + "inferrsi.common.skills.host_memory_management", + "inferrsi.common.skills.bound_windowed_attention_replay" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle" + ], + "audited_contexts": 5, + "output_bytes": 1006115, + "context_bytes": 183756, + "context_omissions": [], + "sha256": "1d3a8d8b3098b7cee4d7a4553e91420adb041dbceb05c975884451da50371c8c", + "artifact": "covered255-05-raw.json.gz" + }, + "cue": { + "rank": 1, + "ids": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle", + "inferrsi.ascend.skills.torchnpu_device_memory_lifecycle", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.host_memory_management", + "inferrsi.common.skills.bound_windowed_attention_replay" + ], + "complete_relevant": [ + "inferrsi.ascend.skills.torchnpu_pinned_host_memory_lifecycle" + ], + "audited_contexts": 5, + "output_bytes": 1006078, + "context_bytes": 183756, + "context_omissions": [], + "sha256": "f744bb7430d2c79786ec30b912af2a84f04c95613b938be82b56fde007c924d9", + "artifact": "covered255-05-cue.json.gz" + } + } + }, + { + "id": "covered255-06", + "identity_control": false, + "arms": { + "raw": { + "rank": 1, + "ids": [ + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.common.skills.checkpoint_recovery_consistency", + "inferrsi.common.skills.performance_ceiling_analysis" + ], + "complete_relevant": [ + "inferrsi.common.skills.weight_loading_optimization" + ], + "audited_contexts": 4, + "output_bytes": 961055, + "context_bytes": 143629, + "context_omissions": [ + { + "ref": "package:inferrsi.common.skills.performance_ceiling_analysis", + "reason": "context_limit", + "required_bytes": 170782, + "available_bytes": 118515 + } + ], + "sha256": "161f56d917c53d40a73de6b9b797622d1daf6f128d9add0eab9ae0d90ab642d1", + "artifact": "covered255-06-raw.json.gz" + }, + "cue": { + "rank": 1, + "ids": [ + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.ascend.skills.cann_attribution_overlap" + ], + "complete_relevant": [ + "inferrsi.common.skills.weight_loading_optimization" + ], + "audited_contexts": 4, + "output_bytes": 945736, + "context_bytes": 129803, + "context_omissions": [ + { + "ref": "package:inferrsi.common.skills.performance_ceiling_analysis", + "reason": "context_limit", + "required_bytes": 170782, + "available_bytes": 141313 + } + ], + "sha256": "4a5366c58b109a22bd160c74664d8cad12aac1502cc13a022f736a0e0d07072c", + "artifact": "covered255-06-cue.json.gz" + } + } + } + ], + "errors": [] +} diff --git a/evidence/indexer-255-covered-cue-20261030/results/run.log b/evidence/indexer-255-covered-cue-20261030/results/run.log new file mode 100644 index 000000000..ed57c4a60 --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/results/run.log @@ -0,0 +1 @@ +{"verdict": "FAIL", "metrics": {"raw": {"hit": 6, "mrr_exact": "29/36", "complete_relevant": 6}, "cue": {"hit": 6, "mrr_exact": "29/36", "complete_relevant": 6}}, "gates": {"no_hit_lost": true, "hit_nondecreasing": true, "mrr_strictly_greater": false, "complete_nondecreasing": true, "identity_control_parity": true, "default_parity": true, "resource_audits": true}, "errors": []} diff --git a/evidence/indexer-255-covered-cue-20261030/validator/evaluate.py b/evidence/indexer-255-covered-cue-20261030/validator/evaluate.py new file mode 100644 index 000000000..b7237912b --- /dev/null +++ b/evidence/indexer-255-covered-cue-20261030/validator/evaluate.py @@ -0,0 +1,120 @@ +"""Frozen single paired CPU retrieval capture; no rewriting or candidate implementation.""" +import gzip +import hashlib +import json +import sys +from fractions import Fraction +from pathlib import Path + +root, prep, output = (Path(x).resolve() for x in sys.argv[1:]) +sys.path.insert(0, str(root / 'src')) +from inferrsi.knowledge_operations import KnowledgeReader +from inferrsi._skill_selection import indexed_skill + + +def encode(value): + return json.dumps(value, ensure_ascii=False, sort_keys=True).encode() + + +def sha(data): + return hashlib.sha256(data).hexdigest() + + +def write(name, value): + (output / name).write_text(json.dumps(value, ensure_ascii=False, indent=2) + '\n') + + +freeze = json.loads((prep / 'input-freeze.json').read_text()) +for path, expected in freeze['files'].items(): + assert sha((prep / path).read_bytes()) == expected +source = json.loads((prep / 'source-manifest.json').read_text())['files'] +assert all(sha((root / path).read_bytes()) == digest for path, digest in source.items()) +cases = json.loads((prep / 'evaluation-inputs.json').read_text())['cases'] +rows = [] +errors = [] +for case in cases: + assert sha((root / case['evidence_path']).read_bytes()) == case['source_sha256'] + pair = {'id': case['id'], 'identity_control': case['identity_control'], 'arms': {}} + for arm, field in (('raw', 'raw_task'), ('cue', 'cue')): + try: + reader = KnowledgeReader() + result = reader.collect(query=case['query'], relevance=case[field], + limit=case['limit'], context_bytes=case['context_bytes'], + **case['filters']) + data = encode(result) + name = case['id'] + '-' + arm + '.json.gz' + (output / name).write_bytes(gzip.compress(data, mtime=0)) + ids = [item['id'] for item in result['items']] + rank = next((i + 1 for i, identity in enumerate(ids) + if identity in case['relevant_ids']), None) + complete = [] + audited = 0 + for context in result['contexts']: + bundle = reader.read(context['ref'], revision=context['revision']) + view = {**bundle.view(), 'authority': 'package'} + _, required, _ = indexed_skill(reader.records()[view['id']], view, bundle.content) + emitted = {r['path']: r for r in context['resources']} + assert all(path in emitted for path in required) + assert all(r['content'].encode() == bundle.content[path] + and r['sha256'] == sha(bundle.content[path]) + for path, r in emitted.items()) + assert context['content_bytes'] == sum(len(r['content'].encode()) + for r in emitted.values()) + audited += 1 + if view['id'] in case['relevant_ids']: + complete.append(view['id']) + assert result['context_bytes'] == sum(c['content_bytes'] for c in result['contexts']) + assert result['context_bytes'] <= case['context_bytes'] + for omission in result['context_omissions']: + if omission['reason'] == 'context_limit': + assert omission['required_bytes'] > omission['available_bytes'] + pair['arms'][arm] = dict(rank=rank, ids=ids, complete_relevant=complete, + audited_contexts=audited, output_bytes=len(data), + context_bytes=result['context_bytes'], + context_omissions=result['context_omissions'], + sha256=sha(data), artifact=name) + except Exception as exc: + errors.append(dict(case=case['id'], arm=arm, type=type(exc).__name__, message=str(exc))) + rows.append(pair) + write('partial-results.json', {'rows': rows, 'errors': errors}) + +captures = [] +for arm in ('raw', 'cue'): + reader = KnowledgeReader() + identity = 'inferrsi.common.skills.cgroup_resource_controls' + values = [] + for options in ({}, {'relevance': 'cgroupv2', 'context_bytes': 262144}, + {'query': 'stat.loc', 'relevance': 'cgroupv2'}, + {'relevance': 'gpu', 'backend': 'ascend'}, + {'ids': [identity], 'relevance': 'ram', 'context_bytes': 262144}, + {'ids': [identity], 'relevance': 'ram', 'context_bytes': 1}): + data = encode(reader.collect(**options)) + values.append({'options': options, 'sha256': sha(data), 'bytes': len(data)}) + captures.append(values) +write('default-captures.json', {'raw': captures[0], 'cue': captures[1]}) +source_same = all(sha((root / path).read_bytes()) == digest for path, digest in source.items()) +metrics = {} +gates = {} +if not errors: + for arm in ('raw', 'cue'): + ranks = [row['arms'][arm]['rank'] for row in rows] + metrics[arm] = {'hit': sum(rank is not None for rank in ranks), + 'mrr_exact': str(sum((Fraction(1, r) if r else Fraction(0) + for r in ranks), Fraction(0)) / len(ranks)), + 'complete_relevant': sum(len(row['arms'][arm]['complete_relevant']) + for row in rows)} + gates = {'no_hit_lost': all(not row['arms']['raw']['rank'] or row['arms']['cue']['rank'] + for row in rows), + 'hit_nondecreasing': metrics['cue']['hit'] >= metrics['raw']['hit'], + 'mrr_strictly_greater': Fraction(metrics['cue']['mrr_exact']) > + Fraction(metrics['raw']['mrr_exact']), + 'complete_nondecreasing': metrics['cue']['complete_relevant'] >= + metrics['raw']['complete_relevant'], + 'identity_control_parity': all(row['arms']['raw']['sha256'] == + row['arms']['cue']['sha256'] for row in rows + if row['identity_control']), + 'default_parity': captures[0] == captures[1], 'resource_audits': True} +verdict = 'INCONCLUSIVE' if errors or not source_same else ('PASS' if all(gates.values()) else 'FAIL') +write('results.json', {'verdict': verdict, 'metrics': metrics, 'gates': gates, + 'source_unchanged': source_same, 'rows': rows, 'errors': errors}) +print(json.dumps({'verdict': verdict, 'metrics': metrics, 'gates': gates, 'errors': errors})) diff --git a/evidence/indexer-255-token-v3-20261030/README.md b/evidence/indexer-255-token-v3-20261030/README.md new file mode 100644 index 000000000..fd134dbc5 --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/README.md @@ -0,0 +1,8 @@ +# Structured string-value tokenization v3: FAIL + +[Report and per-case table](results/report.md). Production baseline b3549ce1; rejected candidate96788d26. MRR drops7/12→8/15 despite Hit4/6→5/6. No runtime promotion, heldout execution or #255 closure. + +Frozen scripts are in validator/, inputs in validator/inputs/. For authorized replay set PREP_ROOT to validator/inputs/, ROOT to the pinned baseline or candidate tree, and create a new OUTPUT_DIR. Candidate patch/bundle is in patch/ (baseline prerequisite b3549ce1). Frozen report uses original artifact basenames; this README maps their packaged locations. The original local results integrity inventory is replaced by plan/publication-integrity.json for this relocated bundle; frozen artifact bytes are unchanged. pre-semantic-freeze.json records its pre-run state, not final execution status. + +### Frozen preparation filename mapping +`plan/source-pins.json` is the exact original source-manifest.json snapshot (not a current-tree hash validator). For replay copy plan/ to a new external preparation directory and copy source-pins.json to source-manifest.json there; do not modify the evidence tree. Internal frozen references retain the original filename and hash. `plan/preparation-freeze.json` is the exact original freeze-manifest.json receipt, likewise renamed only for explicit snapshot classification. Publication-integrity validates both snapshot bytes, not historical corpus availability. diff --git a/evidence/indexer-255-token-v3-20261030/patch/candidate.bundle b/evidence/indexer-255-token-v3-20261030/patch/candidate.bundle new file mode 100644 index 000000000..32b938296 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/patch/candidate.bundle differ diff --git a/evidence/indexer-255-token-v3-20261030/patch/candidate.patch b/evidence/indexer-255-token-v3-20261030/patch/candidate.patch new file mode 100644 index 000000000..6313ef6f1 --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/patch/candidate.patch @@ -0,0 +1,491 @@ +diff --git a/src/inferrsi/_skill_selection.py b/src/inferrsi/_skill_selection.py +index d06f492c..fe81a711 100644 +--- a/src/inferrsi/_skill_selection.py ++++ b/src/inferrsi/_skill_selection.py +@@ -21,6 +21,28 @@ def lexical_score(cue: str, fields: list[tuple[int, object]]) -> int: + ) + + ++def string_value_words(value: object) -> set[str]: ++ """Union tokens from scalar string values, never keys or serialized syntax.""" ++ if isinstance(value, str): ++ return words(value) ++ if isinstance(value, dict): ++ values = value.values() ++ elif isinstance(value, (list, tuple)): ++ values = value ++ else: ++ return set() ++ tokens: set[str] = set() ++ for item in values: ++ tokens.update(string_value_words(item)) ++ return tokens ++ ++ ++def token_score(cue: str, fields: list[tuple[int, object]]) -> int: ++ """Experimental string-value token presence; no morphology or key evidence.""" ++ tokens = words(cue) ++ return sum(weight * len(tokens & string_value_words(value)) for weight, value in fields) ++ ++ + def indexed_skill(record: dict, view: dict, content: dict[str, bytes]) -> tuple: + """Validate the existing projection against this exact canonical bundle.""" + registry_path = "data/skill_library/skill-index.json" +@@ -82,8 +104,13 @@ def indexed_skill(record: dict, view: dict, content: dict[str, bytes]) -> tuple: + return core, required, optional + + +-def skill_score(cue: str, record: dict, core: bytes) -> int: +- return lexical_score(cue, [ ++def skill_score( ++ cue: str, record: dict, core: bytes, *, method: str = "weighted-lexical-v1" ++) -> int: ++ if method not in ("weighted-lexical-v1", "weighted-token-v3"): ++ raise ValueError("Unknown relevance method") ++ scorer = token_score if method == "weighted-token-v3" else lexical_score ++ return scorer(cue, [ + (4, record["title"]), + (3, record["summary"]), + (2, [record.get("tags", []), record.get("applicability", {})]), +diff --git a/src/inferrsi/arguments.py b/src/inferrsi/arguments.py +index 17c614df..58fad4bb 100644 +--- a/src/inferrsi/arguments.py ++++ b/src/inferrsi/arguments.py +@@ -191,6 +191,11 @@ def parser() -> argparse.ArgumentParser: + collect.add_argument("--max-bytes", type=int) + collect.add_argument("--material", action="store_true") + collect.add_argument("--relevance", help="Opt-in lexical cue for packaged skill selection") ++ collect.add_argument( ++ "--relevance-method", choices=("weighted-lexical-v1", "weighted-token-v3"), ++ default="weighted-lexical-v1", ++ help="Skill ranking method; token-v3 is experimental and requires --relevance", ++ ) + collect.add_argument( + "--context-bytes", type=int, + help="Opt-in context view; total UTF-8 resource-body byte budget", +diff --git a/src/inferrsi/cli.py b/src/inferrsi/cli.py +index 9b883658..349e3df3 100644 +--- a/src/inferrsi/cli.py ++++ b/src/inferrsi/cli.py +@@ -193,6 +193,7 @@ def main(argv: list[str] | None = None) -> int: + max_bytes=args.max_bytes, + material=args.material, + relevance=args.relevance, ++ relevance_method=args.relevance_method, + context_bytes=args.context_bytes, + ) + if args.json: +diff --git a/src/inferrsi/knowledge_operations.py b/src/inferrsi/knowledge_operations.py +index 1c78418f..df12561b 100644 +--- a/src/inferrsi/knowledge_operations.py ++++ b/src/inferrsi/knowledge_operations.py +@@ -490,6 +490,7 @@ class KnowledgeReader: + max_bytes: int | None = None, + material: bool = False, + relevance: str | None = None, ++ relevance_method: str = "weighted-lexical-v1", + context_bytes: int | None = None, + ) -> dict: + if type(limit) is not int or limit < 1: +@@ -504,6 +505,10 @@ class KnowledgeReader: + raise ValueError("Unknown content kind") + if relevance is not None and (not isinstance(relevance, str) or not words(relevance)): + raise ValueError("relevance requires at least one word") ++ if relevance_method not in ("weighted-lexical-v1", "weighted-token-v3"): ++ raise ValueError("Unknown relevance method") ++ if relevance_method != "weighted-lexical-v1" and relevance is None: ++ raise ValueError("relevance_method requires a relevance cue") + if context_bytes is not None and (type(context_bytes) is not int or context_bytes < 1): + raise ValueError("context_bytes must be a positive integer") + if revisions is None: +@@ -645,7 +650,9 @@ class KnowledgeReader: + continue + try: + core, _, _ = indexed_skill(record, view, bundle.content) +- score = skill_score(relevance, record, bundle.content[core]) ++ score = skill_score( ++ relevance, record, bundle.content[core], method=relevance_method ++ ) + except (ValueError, KeyError, TypeError) as exc: + omitted.append({ + "id": reference, "reason": "index_unavailable", "detail": str(exc), +@@ -673,7 +680,7 @@ class KnowledgeReader: + } + if relevance is not None: + result["selection"] = { +- "method": "weighted-lexical-v1", "cue": relevance, ++ "method": relevance_method, "cue": relevance, + "scope": "generated-packaged-skills", "ranking_applied": ids is None, + } + result["limitations"][0] = ( +diff --git a/tests/fixtures/indexer/accuracy-token-v3/synthetic.json b/tests/fixtures/indexer/accuracy-token-v3/synthetic.json +new file mode 100644 +index 00000000..32acc90f +--- /dev/null ++++ b/tests/fixtures/indexer/accuracy-token-v3/synthetic.json +@@ -0,0 +1,281 @@ ++{ ++ "status": "prospective before implementation; manually specified expected arithmetic, no candidate evaluated", ++ "cases": [ ++ { ++ "id": "token-dev01", ++ "cue": "ram", ++ "fields": [ ++ [ ++ 4, ++ "program analysis" ++ ], ++ [ ++ 1, ++ "RAM pressure" ++ ] ++ ], ++ "expected_whole_token_score": 1 ++ }, ++ { ++ "id": "token-dev02", ++ "cue": "io", ++ "fields": [ ++ [ ++ 4, ++ "optimization" ++ ], ++ [ ++ 3, ++ "I/O telemetry" ++ ], ++ [ ++ 1, ++ "io tracing" ++ ] ++ ], ++ "expected_whole_token_score": 1 ++ }, ++ { ++ "id": "token-dev03", ++ "cue": "GPU gpu", ++ "fields": [ ++ [ ++ 4, ++ "gpu GPU" ++ ], ++ [ ++ 1, ++ "GPU" ++ ] ++ ], ++ "expected_whole_token_score": 5 ++ }, ++ { ++ "id": "token-dev04", ++ "cue": "cpu.stat.local", ++ "fields": [ ++ [ ++ 4, ++ "cpu stat local" ++ ], ++ [ ++ 1, ++ "cpu.stat.local" ++ ] ++ ], ++ "expected_whole_token_score": 15 ++ }, ++ { ++ "id": "token-dev05", ++ "cue": "read_cache", ++ "fields": [ ++ [ ++ 4, ++ "read cache" ++ ], ++ [ ++ 1, ++ "read_cache" ++ ] ++ ], ++ "expected_whole_token_score": 1 ++ }, ++ { ++ "id": "token-dev06", ++ "cue": "checkpoint", ++ "fields": [ ++ [ ++ 4, ++ "checkpoints" ++ ], ++ [ ++ 1, ++ "checkpoint" ++ ] ++ ], ++ "expected_whole_token_score": 1 ++ }, ++ { ++ "id": "token-dev07", ++ "cue": "\u6570\u636e", ++ "fields": [ ++ [ ++ 4, ++ "\u6570\u636e\u5904\u7406" ++ ], ++ [ ++ 1, ++ "\u6570\u636e" ++ ] ++ ], ++ "expected_whole_token_score": 1 ++ }, ++ { ++ "id": "token-dev08", ++ "cue": "orchid", ++ "fields": [ ++ [ ++ 4, ++ "memory allocator" ++ ], ++ [ ++ 1, ++ "stream ownership" ++ ] ++ ], ++ "expected_whole_token_score": 0 ++ }, ++ { ++ "id": "string-dev09", ++ "cue": "whether next", ++ "fields": [ ++ [ ++ 1, ++ "prefix\nwhether\tnext" ++ ] ++ ], ++ "expected_whole_token_score": 2 ++ }, ++ { ++ "id": "string-dev10", ++ "cue": "whether next", ++ "fields": [ ++ [ ++ 1, ++ "prefix whether next" ++ ] ++ ], ++ "expected_whole_token_score": 2 ++ }, ++ { ++ "id": "string-dev11", ++ "cue": "quote beta", ++ "fields": [ ++ [ ++ 3, ++ "\"quote\" \\ beta" ++ ] ++ ], ++ "expected_whole_token_score": 6 ++ }, ++ { ++ "id": "string-dev12", ++ "cue": "input_contract source", ++ "fields": [ ++ [ ++ 2, ++ { ++ "input_contract": "source" ++ } ++ ] ++ ], ++ "expected_whole_token_score": 2 ++ }, ++ { ++ "id": "string-dev13", ++ "cue": "gpu cpu ram", ++ "fields": [ ++ [ ++ 2, ++ [ ++ "GPU", ++ { ++ "x": [ ++ "CPU", ++ "RAM", ++ "gpu" ++ ] ++ } ++ ] ++ ] ++ ], ++ "expected_whole_token_score": 6 ++ }, ++ { ++ "id": "string-dev14", ++ "cue": "null true 42", ++ "fields": [ ++ [ ++ 2, ++ [ ++ null, ++ true, ++ 42 ++ ] ++ ] ++ ], ++ "expected_whole_token_score": 0 ++ }, ++ { ++ "id": "string-dev15", ++ "cue": "null true 42", ++ "fields": [ ++ [ ++ 2, ++ [ ++ "null", ++ "true", ++ "42" ++ ] ++ ] ++ ], ++ "expected_whole_token_score": 6 ++ }, ++ { ++ "id": "string-dev16", ++ "cue": "download", ++ "fields": [ ++ [ ++ 3, ++ "downloads" ++ ], ++ [ ++ 1, ++ "download" ++ ] ++ ], ++ "expected_whole_token_score": 1 ++ }, ++ { ++ "id": "string-dev17", ++ "cue": "ram", ++ "fields": [ ++ [ ++ 4, ++ "program" ++ ], ++ [ ++ 1, ++ { ++ "ram": "dram" ++ } ++ ] ++ ], ++ "expected_whole_token_score": 0 ++ }, ++ { ++ "id": "string-dev18", ++ "cue": "read_cache", ++ "fields": [ ++ [ ++ 1, ++ "read_cache" ++ ], ++ [ ++ 4, ++ "read cache" ++ ] ++ ], ++ "expected_whole_token_score": 1 ++ } ++ ], ++ "unchanged_contracts": [ ++ "default mode byte parity", ++ "explicit ID order/membership/revision", ++ "query substring AND matching", ++ "filters before ranking", ++ "mandatory-resource byte/hash/revision and budget omission", ++ "legacy optional-resource scorer", ++ "qualified-ID deterministic tie order" ++ ] ++} +diff --git a/tests/test_collect_token_selection.py b/tests/test_collect_token_selection.py +new file mode 100644 +index 00000000..41a3317e +--- /dev/null ++++ b/tests/test_collect_token_selection.py +@@ -0,0 +1,78 @@ ++import json ++from pathlib import Path ++ ++import pytest ++ ++from inferrsi._skill_selection import token_score ++from inferrsi.knowledge_operations import KnowledgeReader ++ ++CASES = json.loads((Path(__file__).parent / 'fixtures/indexer/accuracy-token-v3/synthetic.json') ++ .read_text())['cases'] ++ ++ ++@pytest.mark.parametrize('case', CASES, ids=lambda case: case['id']) ++def test_frozen_token_contract(case): ++ assert token_score(case['cue'], case['fields']) == case['expected_whole_token_score'] ++ ++ ++def test_default_and_explicit_default_identical(): ++ reader = KnowledgeReader() ++ for options in ({}, {'relevance': 'candidate', 'context_bytes': 262144}): ++ assert reader.collect(**options) == reader.collect( ++ **options, relevance_method='weighted-lexical-v1') ++ ++ ++def test_opt_in_validation(): ++ reader = KnowledgeReader() ++ with pytest.raises(ValueError, match='requires a relevance'): ++ reader.collect(relevance_method='weighted-token-v3') ++ with pytest.raises(ValueError, match='Unknown relevance'): ++ reader.collect(relevance='gpu', relevance_method='unknown') ++ ++ ++def test_explicit_selection_resources_and_budget_unchanged(): ++ reader = KnowledgeReader() ++ identity = 'inferrsi.common.skills.cgroup_resource_controls' ++ revision = reader.read(identity).revision ++ for budget in (1, 262144): ++ options = dict(ids=[identity], revisions={identity: revision}, relevance='ram', ++ context_bytes=budget) ++ legacy = reader.collect(**options) ++ token = reader.collect(**options, relevance_method='weighted-token-v3') ++ assert token['selection']['ranking_applied'] is False ++ token['selection']['method'] = legacy['selection']['method'] ++ assert token == legacy ++ ++ ++def test_token_operational_controls_and_substring_query(): ++ reader = KnowledgeReader() ++ options = dict(relevance_method='weighted-token-v3') ++ assert reader.collect(relevance='zzzznonexistent255token', **options)['no_match'] ++ assert reader.collect(relevance='cgroupv2', repo='not-a-real-repo', **options)['no_match'] ++ result = reader.collect('stat.loc', relevance='cgroupv2', **options) ++ assert result['items'][0]['id'] == 'inferrsi.common.skills.cgroup_resource_controls' ++ assert reader.collect('stat.loc zzzznonexistent255token', relevance='cgroupv2', ++ **options)['no_match'] ++ ++ ++def test_token_ties_use_qualified_identity_order(): ++ from inferrsi._skill_selection import skill_score ++ reader = KnowledgeReader() ++ result = reader.collect(relevance='candidate', relevance_method='weighted-token-v3', limit=100) ++ ordered = [] ++ for item in result['items']: ++ bundle = reader.read(item['ref']) ++ record = reader.records()[item['id']] ++ core = record['skill_documents']['original'] ++ ordered.append((-skill_score('candidate', record, bundle.content[core], ++ method='weighted-token-v3'), item['ref'])) ++ assert ordered == sorted(ordered) ++ ++ ++def test_cli_token_method(capsys): ++ from inferrsi.cli import main ++ assert main(['collect', '', '--relevance', 'cgroupv2', '--relevance-method', ++ 'weighted-token-v3', '--json']) == 0 ++ result = json.loads(capsys.readouterr().out) ++ assert result['selection']['method'] == 'weighted-token-v3' ++ assert result['items'][0]['id'] == 'inferrsi.common.skills.cgroup_resource_controls' diff --git a/evidence/indexer-255-token-v3-20261030/plan/pre-semantic-freeze.json b/evidence/indexer-255-token-v3-20261030/plan/pre-semantic-freeze.json new file mode 100644 index 000000000..51efa0ed4 --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/plan/pre-semantic-freeze.json @@ -0,0 +1,94 @@ +{ + "candidate": "96788d26b57240e53c26fe5446b2062a62ee6d0e", + "tree": "5ae435fb5793b19f0de96ce74dd6ded9db3f3292", + "baseline": "b3549ce1aea1556f76de1d2e14a4240474947b6b", + "source": { + "src/inferrsi/__init__.py": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47", + "src/inferrsi/__main__.py": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d", + "src/inferrsi/_skill_selection.py": "a5846a1077f7ecdf36f4e93f58247f7cf7d4d14aa5b45dd21d2d58f696e9c5b3", + "src/inferrsi/arguments.py": "559c1da102ca69e8809fe9001a979be083f7a5a8ea5086eba0dd71c4c4c39c49", + "src/inferrsi/catalog.py": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5", + "src/inferrsi/claim.py": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1", + "src/inferrsi/cli.py": "2937c0ce2fd218244b2350cd1d5cd6b85ec25ed98a58262f159ae3f8c85b745c", + "src/inferrsi/contracts.py": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4", + "src/inferrsi/data/skill_library/methods/video-diffusion-optimization/scripts/evaluate_target.py": "9e84df051e03d567746fd73ae7243efb2f3c5ec806fef824d3c4d42a8eb9704c", + "src/inferrsi/data/skill_library/repositories/dwarfstar/scripts/request_template.py": "62e4cde002f2aebeaddd63e6e1f04fc727c271d2cbdaa00856158baadee5340b", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/scripts/audit_fasth3.py": "05bfb9348cf88a76f4fba8bc8a600fe54246d83a98c1685bc4847ace28e2fbac", + "src/inferrsi/data/skill_library/repositories/husky/scripts/audit_evidence_card.py": "291e62701ef7a37a5bfab963ea4e20abb94245bc015584223a101f18b1490895", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/model-runtime/scripts/screen_config.py": "5961bb6f31ceca43473683127b37faee268234040a013ae49bbb4898eacb6f04", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/qualification/scripts/audit_benchmark_record.py": "59534f58d46ad438c7bf88886a9b05a8716c922d817f2d9c58ff1b71ae50ee85", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/serving-cache/scripts/request_template.py": "d6b00c9f0b38c7fb0a0e52e396993a832542ea0aa6e2456026fabc0c39a4d33a", + "src/inferrsi/data/skill_library/repositories/lithos-metal/scripts/inspect_contract.py": "f8365f87af2135fcf0a2c8baaefac563c30989282fe080a02d958a37a215515c", + "src/inferrsi/data/skill_library/repositories/local-dream/scripts/screen_declarations.py": "ccd7d0d55be9a47484ec3a74498d8edef8aba55287f92287775c8d82050c8932", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/scripts/audit_decode_metadata.py": "de28f586f27278fee25f6c96f1e00ee63528cdf746f57c83bdf367c31c0f6e6b", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/scripts/screen_prefill_metadata.py": "769366c19d621ec1b42c722ed224fa154a63c3410a0cffc9072a13abd9c71bbd", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/scripts/warmup_shape_inventory.py": "6aa920ca1cb534dec41f5dd2d833c3e760b234ab2c6439b6c4250b8a30b74055", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/scripts/audit_artifact_declarations.py": "874a16cb16d75a14cb51b4fbe442f4fac46feb21ac242063f5f845a0fcbabaa4", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/scripts/audit_apple_target_declarations.py": "d5bf16f7def4cfe28b143a700ca45a9636ddaeac12ead04cbaecdcc8b4b71aa5", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/scripts/audit_kv_limit_declarations.py": "d82a6dda5f5931eb369003a5e1f419bfbdd82304766a18be8ea1e1849e26b77d", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/scripts/audit_gated_delta_declarations.py": "3a256c37c867866e49dfdc2893cbb3f32d061e389fe4395555eff899bc0415ff", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/scripts/audit_generation_evidence.py": "2aba6caf42ca1b3fc55299fd96419af8681c37750c21a7ba75205a43baa9c7b1", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/scripts/audit_runtime_plan.py": "b096ec10c4f0e4b057997fe7b946007ab4c06dd263ab09d803e29e498601da00", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/qualification/scripts/summarize_decode_records.py": "cc0ac3c403aba79a131f631290d68e1640283e9f0d564aeef431b577824e3036", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/scripts/request_template.py": "d91b03fbbd7003012a92bd7259d39604b9a91ee5707da2d3366e7c0fb47edd6a", + "src/inferrsi/data/skill_library/repositories/splash/scripts/request_template.py": "086d773d42edf67c43cedae16974061b09ac66458ee47fe961fc01733f3b2a0e", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/qualification/scripts/summarize_rounds.py": "107a844954e3909500c60eb1ca365ac9b5e7dc3d91122dbae46f402a5cef4e3b", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/runtime-planning/scripts/parse_budget.py": "b3a3b0847280333c2bfbe5d03350dc866307d5290160ec64c7e3d462a34f6301", + "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/scripts/screen_pool_budget.py": "dd615f7d3239f48510ce83416fb14fba8d5da2b9017627bdd312567be1bd7956", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/scripts/check_environment_plan.py": "2603200fef87e3afad7820b84994f7e43e6041b45c17d0ed613d0d6229535bf2", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/scripts/summarize_serving_records.py": "78e660a13a6a6941d006cb11fc2f401a34f489e54273860ae6a208dca6a6a3c2", + "src/inferrsi/evaluation.py": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e", + "src/inferrsi/evidence_paths.py": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa", + "src/inferrsi/execution.py": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92", + "src/inferrsi/harness/__init__.py": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5", + "src/inferrsi/harness/evaluation.py": "20d0c52145600a5d6b1e513b61f2fffaa459b83db194876be40726bd4f343a59", + "src/inferrsi/harness/improvement.py": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8", + "src/inferrsi/harness/reporting.py": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3", + "src/inferrsi/harness/runner.py": "d290fc87710e77c8e6069e95bb0d347016a393707592d2eed98f07951f2d0865", + "src/inferrsi/harness/scenarios/__init__.py": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a", + "src/inferrsi/harness/scenarios/external.py": "f3f8d4fc516ebce81c26d82248e83f79b299eed0fc17ce73e2844995aea9c3f9", + "src/inferrsi/harness/scenarios/media.py": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a", + "src/inferrsi/harness/skill_eval_diagnostic.py": "e25cfee943052aaefbddf3f545fe6d56d95ab1edb9723cd9353feeabc954ea9a", + "src/inferrsi/harness/skill_eval_engineering.py": "2f18a1ab96acfc84319eba05e94c2a073fd0dca83e4f82ad98142ba2fecda039", + "src/inferrsi/harness/skill_evaluation.py": "ca141c3d1272591e24797d1ee31bed3aa37d9a009897c88c2ce0b087f3da11ac", + "src/inferrsi/infra/__init__.py": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c", + "src/inferrsi/infra/artifacts.py": "bcf3adf325e34e10a4ecb4baf0af855ddca3d77eff76de5b66e493baaa3d7eaf", + "src/inferrsi/infra/campaign.py": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209", + "src/inferrsi/infra/commands.py": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1", + "src/inferrsi/infra/environment.py": "23a6716089c542617761b5a3fc67dde0af9e483212fc003f8577757fbc4ed2db", + "src/inferrsi/infra/executors/__init__.py": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a", + "src/inferrsi/infra/executors/local.py": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77", + "src/inferrsi/infra/http.py": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2", + "src/inferrsi/infra/inspection.py": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd", + "src/inferrsi/infra/platforms/__init__.py": "dac68eb6edbff995f8afe477396d81efd5db93128978ceedaea1babf0dd3db08", + "src/inferrsi/infra/platforms/cuda.py": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f", + "src/inferrsi/infra/platforms/mlx.py": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3", + "src/inferrsi/infra/processes.py": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189", + "src/inferrsi/infra/targets/__init__.py": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3", + "src/inferrsi/infra/targets/git.py": "abb93f9332db336298102e28c06b9a7553597d4d03b957358602f60d61b1aa65", + "src/inferrsi/inspection.py": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3", + "src/inferrsi/knowledge_operations.py": "087f24a8bf0aba7375702eaddd7ce582a03303644a712dd0333dd0cae4366793", + "src/inferrsi/knowledge_store.py": "70dfca059b0a682579206d4376d191667934b3ca869ae847098c4c6964d86abe", + "src/inferrsi/model/__init__.py": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f", + "src/inferrsi/model/anthropic_agent.py": "746db742467fee153427cc9d7e8a01b640bc13e5b9cd431e935f596a8e19b072", + "src/inferrsi/model/protocols/__init__.py": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47", + "src/inferrsi/model/protocols/image.py": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f", + "src/inferrsi/model/protocols/metrics.py": "5b32e296a746b6de30cb0c89f23f730296b0087884c1acc2b6127c1295d4db7a", + "src/inferrsi/model/request_guard.py": "c2f8a2f822f32785327a2138ee7b0609896b180819d9d1cbe851534aa6198628", + "src/inferrsi/model/validators/__init__.py": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b", + "src/inferrsi/model/validators/flare.py": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f", + "src/inferrsi/model/validators/image.py": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a", + "src/inferrsi/reporting.py": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d", + "src/inferrsi/research.py": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04", + "src/inferrsi/runspec.py": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709", + "src/inferrsi/skill_catalog.py": "dbbe107a105e13b0c288f48ad80f9b05e108606af379d243dfab85cdfc99cfb5", + "src/inferrsi/summary_chart.py": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc" + }, + "runners": { + "compatibility_capture.py": "46937bfbb7f2bb2087c8c89d9435a326557fbd0b0c5a25ac5151d20b10d87b40", + "evaluate_development.py": "47570bda5dd642c9a7170dcd24680806a007ecc88432569af9afa4867dbc7db8" + }, + "semantic_runs_started": false +} diff --git a/evidence/indexer-255-token-v3-20261030/plan/preparation-freeze.json b/evidence/indexer-255-token-v3-20261030/plan/preparation-freeze.json new file mode 100644 index 000000000..27bb6699c --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/plan/preparation-freeze.json @@ -0,0 +1,10 @@ +{ + "frozen_at_utc": "2026-10-30T02:56:05.815592+00:00", + "status": "FROZEN_PROSPECTIVE_BEFORE_IMPLEMENTATION_AND_EVALUATION", + "baseline": "b3549ce1aea1556f76de1d2e14a4240474947b6b", + "files": { + "protocol.md": "2adf6a226b4c3763e52dec159aaae9ec4ea0824f2b0b0bd37a92799ef06bf94d", + "source-manifest.json": "9a1a7d9fda980602ca344a4cf8be10fbfa7eef96c2dc1feaea8eab75647c4215", + "synthetic-contracts.json": "4cd0fd6fd7ac9dcb910cd31f560c0507fa2e5701b4bf9110bfb77ef9acfb9dc8" + } +} diff --git a/evidence/indexer-255-token-v3-20261030/plan/protocol.md b/evidence/indexer-255-token-v3-20261030/plan/protocol.md new file mode 100644 index 000000000..6b95352fe --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/plan/protocol.md @@ -0,0 +1,35 @@ +# #255 prospective single-candidate v3 protocol + +Frozen before implementation/evaluation on 2026-10-30 UTC. Owner explicitly requested indexer continuation (02:53); coordinator delegated this bounded trial. This is a new experiment, not a retry/relabel of failed v2. Prior v2 FAIL and all prior efficiency results remain unchanged. + +## Question / exact candidate +Does tokenizing actual structured-field scalar strings instead of JSON serialization avoid boundary/schema artifacts while meeting the original retrieval gates? + +Baseline production: `b3549ce1aea1556f76de1d2e14a4240474947b6b`. Production src is byte-identical to previous1147; source-manifest.json pins 813 src files. Corpus and metadata must not change. Candidate method `weighted-token-v3` is explicitly opt-in; existing weighted-lexical-v1 and unspecified/default calls remain byte-identical. Implement exactly one fixed candidate: + +- Query tokens: unchanged Unicode `re.findall(r"\w+", cue.casefold())` unique set. +- For each existing weighted field, recursively visit scalar string VALUES only: a string yields its tokens; list/tuple elements recurse; dict values recurse; keys and non-string scalars (numbers, booleans, null) contribute nothing. Tokens are unioned within each field, so repeated values do not increase score. Do not concatenate distinct leaves before tokenization. +- Preserve field weights title4, summary3, [tags, applicability]2, core1; weighted query/field set intersection. Decode actual core UTF-8 string without JSON encoding; no special newline substitution list. +- Preserve filtering, substring-AND query, explicit IDs/order/revision, score-positive inclusion and ascending-ref ties, optional-resource lexical scorer, mandatory resource/budget semantics. +- No morphology/stemming/synonyms/stopwords/term rarity/boosts, no corpus edits or model/service calls. Dictionary-key removal is an explicit policy consequence of values-only semantics, tested prospectively, not silently presented as solely whitespace repair. + +## Inputs and independence +Six prior semantic DEV positives are reused TRAINING DIAGNOSTICS: previously observed and used to motivate this hypothesis, not fresh validation. Two prior out-of-domain cases remain nongating diagnostics, not labelled false positives. Eighteen synthetic contracts (prior8 +10 structured/boundary contrasts) and seven unchanged operational contracts are frozen separately. Lexical-absence and incompatible-filter no-match gates remain distinct from domain diagnostics. + +Six sealed source-reviewed positives remain unused for retrieval evaluation. Original source reviewers and case author accessed labels; candidate implementer explicitly confirmed on 2026-10-30 ~02:54 UTC that they have never opened/read their queries/labels or received contents/outcomes (only hash/count and aggregate preparation descriptions). Independent reviewer has never run these cases. This is an access attestation and retained task-history check, not an OS access-log proof. Heldout content must stay out of implementer messages/artifacts until the one frozen independent evaluation, if allowed below. No new selection based on v3 outputs. + +Original semantic manifest SHA256 `a4d6d0d18c3bdb4fbbed887a33fe6cbc07ca1b705650018e8ef13a27d527dd81`. +DEV SHA256 `02b03af2319c48336d103d6ea88b7bc7e96baadd6de034b52f62f17773e1cad5`. +Heldout SHA256 `ac54bd7b55687bce8d7c6727e3d63b7fe184f5de84b3300bb4f1f9e49baa9500`. +Domain diagnostic SHA256 `f624283f7f4305d7067ad1cf4e115ce512b8dba2b6459af6458ce4e5f8e308b2`. + +## Prospective gates and order +1. Freeze protocol/input hashes before implementation. Freeze candidate commit/source hashes and evaluation runner before semantic retrieval. One implementation only; repairs needed after results invalidate this candidate rather than allowing tune-until-pass. +2. Synthetic18/18 and unchanged contracts7/7; separately lexical-absence/incompatible-filter no-match; exact serialized default-mode parity vs production baseline on retained six parity scenarios. No default/runtime promotion. +3. Run six DEV paired against production baseline, one time. Hit@5 AND MRR@5 must be no lower than production baseline. Keep the same gates as v2, not a weaker comparator chosen to pass. Historical v2 frozen DEV is reported in a separate diagnostic column (no new v2 run, no pooling; same corpus). V3 must not merely beat failed v2. All complete-resource/budget audits below must pass. If DEV fails, STOP before heldout; retain FAIL, no tuning or retries. +4. Only if all prior gates pass: independent reviewer (not implementer) runs six heldout once paired vs production baseline. Hit@5 must gain at least ONE labelled positive hit; no baseline positive hit may be lost; MRR@5 nondecreasing. Do not publish/release labels for tuning or rerun after seeing outcomes. +5. Every emitted context must carry its full indexed_skill required set, byte/hash/revision fidelity and proper omission/budget accounting. Relevant complete delivery count must not decrease vs baseline. Minimum case suffixes are anchors only, not exhaustive coverage. Preserve raw outputs, exclusions and failures; no hidden dropped cases. +6. Full passed gates yield only bounded small-set evidence suitable for independent code/CI review; not automatic production promotion, merge, efficacy claim or #255 closure. Failure = FAIL; missing evidence/identity drift = INCONCLUSIVE, not pass. No latency/performance conclusions from this untimed CPU retrieval trial. + +## Limits / reporting +Report per-case labelled rank, Hit@5/MRR@5, complete relevant deliveries, context bytes and omissions. Partial positives do not establish population accuracy, complete precision/recall, downstream agent benefit or hardware efficacy. Reused DEV results are hypothesis-development evidence. Historical v2 FAIL is never overwritten. If heldout is accessed/evaluated prematurely or exposed to implementer, stop and report independence compromised rather than silently replacing cases. No installations, model/API/hardware or paid experiments; no broad benchmark or new service. Any further candidate requires a separate prospective decision, not this authorization. diff --git a/evidence/indexer-255-token-v3-20261030/plan/publication-integrity.json b/evidence/indexer-255-token-v3-20261030/plan/publication-integrity.json new file mode 100644 index 000000000..fe76b4307 --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/plan/publication-integrity.json @@ -0,0 +1,37 @@ +{ + "README.md": "818f87248ac9446d230cb31d24a580026f04b68ae8209104390659b447e85eae", + "patch/candidate.bundle": "1b672634b5578f2b5b31878cefa1eb89edf9edf3deeb64c1b04a88beda438289", + "patch/candidate.patch": "245262a328100796e58a6cc27b3d2e46b516d77f7123543242abef0951cf7da2", + "plan/pre-semantic-freeze.json": "98883caf2e4003d43f2505ededc3bec6a3174997b9c53a7a49f82ef4ea848d1c", + "plan/preparation-freeze.json": "fe23657ea2ffe4cd13fd8f9d5d4c19690f1c9b2d6924486e193a825f136c2e0d", + "plan/protocol.md": "2adf6a226b4c3763e52dec159aaae9ec4ea0824f2b0b0bd37a92799ef06bf94d", + "plan/source-pins.json": "9a1a7d9fda980602ca344a4cf8be10fbfa7eef96c2dc1feaea8eab75647c4215", + "plan/synthetic-contracts.json": "4cd0fd6fd7ac9dcb910cd31f560c0507fa2e5701b4bf9110bfb77ef9acfb9dc8", + "results/baseline-default.json": "f238329d5bc1d8cbf549ddb2b3637840a2bf4c057e8ba25b4c57861abdfcbede", + "results/baseline-freshdev01.json.gz": "9683515a0a90125edbc2a494ff9d97e03654eb7700695a58b61f17fb5414aafd", + "results/baseline-freshdev02.json.gz": "3ac8771f3e40e14eb9da47cd1284c683491f8ee4968a89770395d3b9c90a6c49", + "results/baseline-freshdev03.json.gz": "24df4affd3879986cb4089c84243d8f3f7afbddf1038e1d502de751029b832b9", + "results/baseline-freshdev04.json.gz": "9fb82c97f1ac3f67abef7706f50118815cf35e6b14eab82ab9be6b05cd53b3ff", + "results/baseline-freshdev05.json.gz": "c2211daf1ffabfff0e90c3bad2e6e6f5a84cf53c9eb0b2b14cdb9380f0c5a543", + "results/baseline-freshdev06.json.gz": "4bec8b9ce9077248e8f002bb13567664df71168ea37688abfed372ca5a24307d", + "results/baseline-freshnegative01.json.gz": "7b70948a577c3aa8f430718309df9ce8e8a4ab8ef7054472f74dd173fe725bab", + "results/baseline-freshnegative02.json.gz": "4b96926e945b8b31f968e786e6e5a618c3101922a7e5f829e401833fed09f8c6", + "results/baseline-summary.json": "acb20ae7397617687ca3e150b41c42fd906decb6e8e6d76e2f73b87f73d2073f", + "results/candidate-default.json": "f238329d5bc1d8cbf549ddb2b3637840a2bf4c057e8ba25b4c57861abdfcbede", + "results/compatibility-tests.txt": "6b238b7419b7e8e3d9ce45e52753f35460269a075006331c21a43d7b5b689928", + "results/independent-review.json": "9145af781999b4ec3e0471a3210147cf245c0859993b35e2f4a227121662009f", + "results/report.md": "b58be2e52508d425ab5e0b34bf2bdb107bb32d8caff8749f1d194a16b520be8a", + "results/token-freshdev01.json.gz": "3f4a5484715035baeadd92a0e64f7d9402dafff9eccf84a3dff517c17a390bb5", + "results/token-freshdev02.json.gz": "428c80858059622c5fe654a5ec41304510186c40aeda461917b23d2a2fef72ee", + "results/token-freshdev03.json.gz": "df04a606910247ad3303ca0dc6139ee8111bfd2dd7a9b4568387b679d7271899", + "results/token-freshdev04.json.gz": "38dab17bc5e0ee8b848c45f533fb94090863844794b75908da94da2871a2dfb4", + "results/token-freshdev05.json.gz": "5fa0d8aaade3ccf76b860786fdb0957d77c585e950c7d1b34fa906bb82beb900", + "results/token-freshdev06.json.gz": "67844db80222a1cb5827bc9e16ccade2643ece100f9700b84f75bf227780cbcb", + "results/token-freshnegative01.json.gz": "43e35e13109b1b18514a43e72a44ff447b965fe9dd6405045eaf2b5453f676a3", + "results/token-freshnegative02.json.gz": "9a86fbcd273dbfd50b8d232684145457cad9083988a70b530f9bb2789147e28e", + "results/token-summary.json": "5754ed69e9962a5d03ae8a231da955a5765f177ad2bea760f326c5456f1433eb", + "validator/compatibility_capture.py": "46937bfbb7f2bb2087c8c89d9435a326557fbd0b0c5a25ac5151d20b10d87b40", + "validator/evaluate_development.py": "47570bda5dd642c9a7170dcd24680806a007ecc88432569af9afa4867dbc7db8", + "validator/inputs/semantic-cases/development.json": "02b03af2319c48336d103d6ea88b7bc7e96baadd6de034b52f62f17773e1cad5", + "validator/inputs/semantic-cases/negative-controls.json": "f624283f7f4305d7067ad1cf4e115ce512b8dba2b6459af6458ce4e5f8e308b2" +} diff --git a/evidence/indexer-255-token-v3-20261030/plan/source-pins.json b/evidence/indexer-255-token-v3-20261030/plan/source-pins.json new file mode 100644 index 000000000..32adc2acd --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/plan/source-pins.json @@ -0,0 +1,818 @@ +{ + "commit": "b3549ce1aea1556f76de1d2e14a4240474947b6b", + "files": { + "src/inferrsi/__init__.py": "606691627d4b346e60072208ada8d6b4df2409de190daffb7795c2cc24517e47", + "src/inferrsi/__main__.py": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d", + "src/inferrsi/_skill_selection.py": "4640399f4013425589f6702e2db844820619ba35967d1c3519978eb307168440", + "src/inferrsi/arguments.py": "4456600cf4a362cf069594fe36db46932cf5cb14ddd37683aaf3538f5cf76b4d", + "src/inferrsi/catalog.py": "da5d5b01fed6a4a7027c7249a5514c7fdefbeabd9911846add1f69ae3ce3fbe5", + "src/inferrsi/claim.py": "5f24c4fa91b48f0bb4d4554531607763fc8fc108e90f332816f06cfb520b5cc1", + "src/inferrsi/cli.py": "0bb83bce32171074506a15ea3ea0617456ad1477b53fd568dafb449f88877591", + "src/inferrsi/contracts.py": "03000cb9246ce8c0967147533f7ad25df918ba2674839801453ffb870f496490", + "src/inferrsi/data/knowledge_packs/README.md": "b877e67d9cccf0dcbe57419eeb3ff9374e5bd03ff8eec63adef1025c446140e9", + "src/inferrsi/data/knowledge_packs/dependency/diffusers/cuda/diffusion/group-offload-transfer-overlap.json": "9bec3d912192efa2f2c853d3fb70bdae21a75875630a02294fcd084b18c30414", + "src/inferrsi/data/knowledge_packs/dependency/flashattention/cuda/attention/generation-specific-dispatch.json": "90290d7bc80b1626863e0520c91bb196d93ebfe08190a450e9de59c6505ae8ad", + "src/inferrsi/data/knowledge_packs/dependency/flashinfer/cuda/llm/paged-decode-dispatch.json": "0f2986494fcc9b751df71e598fd61caa68e28bc4b0ac401e4cc763bc3f75283b", + "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/cross-stream-dependency.json": "b829315e55adf8e11f0bd581d5f45dd130d7eb3f81ce7de532f63c7b9c27cbb5", + "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/evaluation-completion.json": "19d2e506ac4f50895776a83a062059aa7a052175c8e4b21dd0d46e1a5a6716a1", + "src/inferrsi/data/knowledge_packs/dependency/mlx/mlx/execution/post-detachment-buffer-lifetime.json": "18d78c0e542dacb42ca33e5264d92aef80fa06d163fd16d4feafd5ed0827ab99", + "src/inferrsi/data/knowledge_packs/dependency/pytorch/cuda/compiler/cudagraph-shapes-and-lifetimes.json": "7fef4122ff794967e61f55d02a4a6b558d7e498d9513607f5ca86db6386928a8", + "src/inferrsi/data/knowledge_packs/dependency/transformers/common/llm/kv-cache-strategy-contracts.json": "a2df26052112a9f3fc6242b2bd636ab13991442db57e56646168d6c52032d38d", + "src/inferrsi/data/knowledge_packs/dependency/triton/common/kernel/bandwidth-fusion-resource-limits.json": "6e9673d1a600a6ad34820e4325f2a6417f44f0f96d8ed0bcf21466d998f19ee4", + "src/inferrsi/data/knowledge_packs/peer/cache_dit/common/diffusion/residual-cache-calibration.json": "bb9509ad0225a91b52bc6787477581d05f0ecd5e44fa59f08f718eea36a2ac45", + "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/bounded-windowed-replay.json": "43613225e9ccdbd34c0f7e3e0767887b9bbfd0f38b2c21684677d99e31a432f0", + "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/speculative-verification-budget.json": "a729e1fc00be44e1480a5dc65bbcdd58df8599681df74dbc651a5a49f27e9028", + "src/inferrsi/data/knowledge_packs/peer/deepinfer/common/llm/zero-overhead-deferred-scheduling.json": "ed5426af145f066f6183bd316c4073c46f324f5ee83aa9bd616387c0d0e7149c", + "src/inferrsi/data/knowledge_packs/peer/lmdeploy/cuda/llm/kv-quantization-capacity-cost.json": "873997b38981ccc03ff73e6fb84e96022f293a019843efd8c4aa50f27878ac9a", + "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/codec-constants-and-state.json": "1cecd7905fef5af3ef8952a4d803900a7df91d662ebae34db47cc0dbf352d4e7", + "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/fixed-customvoice-prompt.json": "ea5e8fdf54cb731b255b73aa0cf47ef53fac103baa2bd5246be45f95d54cb593", + "src/inferrsi/data/knowledge_packs/peer/qwen3_tts_engine/ascend/audio/graph-batching-state-ownership.json": "130741a9f7ac9d366784d6c4dd42860262235caff0c277de1e8a8f8ec933b37f", + "src/inferrsi/data/knowledge_packs/peer/sglang/cuda/llm/hierarchical-prefix-cache.json": "b50ca58df28d4343545d07adf0da23f8f181297b30aba7870e5690630e7d7315", + "src/inferrsi/data/knowledge_packs/peer/tensorrt_llm/cuda/llm/prefill-decode-disaggregation.json": "349856dd98628f94987e014889cd1b15d0e3ae6ab6c576cdde917f28fbafb102", + "src/inferrsi/data/knowledge_packs/peer/xdit/cuda/diffusion/parallelism-freshness-and-topology.json": "3674b5916cb7b384951bf3e449b6796b161878daabcc1d7e7f6d6f8ff861cc17", + "src/inferrsi/data/repo_packs/afd_plugin/pack.json": "00042abeabdc9c0e24f1a4edaba47e233df398aef0ad4e181fc70191d383da5f", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/README.md": "1e28e27d355a66291c665e6096e9a37427d719ef85da8007c59aa292855ccf50", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-aclgraph-capture-replay.json": "53e7846ed11855bde8fc7839d3a4e0ffe3a745841857b2db0ed7c3b6e055e371", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-device-health.json": "06bdc3b9fe83698f0fb504c66b0fbb1105c882075807aea1e2c44bfafd9da8b4", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-distributed-parallel.json": "7dfe1635ee12634a38850004b4b99d46133b89cab7a4f7b948551e2fa3ec36e5", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-graph-mode.json": "173bbac31f9d211a36901f91f7aa3b3898b9e23f7065334226152a04b97b5c99", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-model-adaptation.json": "00de5e35635d25d4e5ba1d04da3f8872e6a23539dec4c9afb32ef816576b58c3", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-profiling.json": "28b3ad941b0b1fc147189b0feb6ac7758f6db1c06a562d1610721c647db3ee84", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-operator-tuning.json": "18f4f5c71a7a15a981f03542aa9a9aab73ed66ca05d9e0edd25c97f60aabf74a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-optimization-casebook.json": "758ecf18433f9e8a74327d158a9e76d8c66511c0effd5ee5472d1271275fbb7a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-performance-analysis.json": "7b4affd23ff614103b41f6d69517f85b97dc80f8f05c4b0f7dd9d4ad4804ea64", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascend-video-diffusion-optimization.json": "a3c31f9a6310fb2f77a350e048f94e49bd6bb742e6e867d6f5ff2701176accbc", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/ascendc-tiling-design.json": "fddaefa64b1b50e8a8f483855fd374740b24aab855a1edc189f095cf8481a3b8", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/cann-attribution-overlap.json": "54f035fb3eb0676ba879cc623aa36fc0133a3b00a7c030cbcc0dcb3a9a49e49c", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-device-memory-lifecycle.json": "affb688fff9a1e20f7cb6bf279634cba4be7ca85ec3f6592647ce6084fd95c08", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/ascend/skills/torchnpu-pinned-host-memory-lifecycle.json": "fd3a156d6d83384f4c93766dc0230e5996f370ad8b0d29257d44055a67454f9d", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-serving-preemption.json": "044b0104345e2cf75e7a8cd801d0ef9dbbdabcd98df3ce067afdd1a97074615e", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/assess-split-backward-schedule.json": "c08af7f6af7b8f42c1fa5dfab987ce8aea0adf07b5cdd86a21f33fd8afbe49bc", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-fixed-workload-claim.json": "14a6d67f478ad8ef9041f845f964b9f7b0d5982c858e0e41d6a437fe23980e7e", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-metal-fasth3-semantics.json": "d49322f8c8c55c89d3651b34a863d4883f1254120e5ce42955f7a2ab23906eb9", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-h3-runtime-residency.json": "bf618f85e546ddc34156e16eb9a40340ab93266c46544584f6408a23087bb8ed", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-llama-runtime-state.json": "c3aa226f04ca06beec17464b719da88f7d89e9dedb8482ec817b77198e3a12df", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-recurrent-kv-finalization.json": "8285bc4fd023c4059a50be83106fb634c8befa32aa1dcc0ff4a11c04ca3e5739", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sd-cpp-residency.json": "dfd72f7b621b3dbb2b983a9c550609de5e7a8aed90e5538fd2b3ddbdeff58513", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-sglang-diffusion-lifetimes.json": "a9df8cbe6650b48f82a52be58613b5010d12e8dd622b2cf9c0961a2f5a98a15a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/audit-vllm-kv-lifetimes.json": "14149492af5e18ecd332e2b078c0d920e3cd992f77baf8a954d7f2fce85ffaf7", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/bound-windowed-attention-replay.json": "8ddd64d11c6a3dcff5678d07475cef499eb823af95ca939e8b25265c57d751c4", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/budget-speculative-verification.json": "7a1a86ca6671e40f5cd3baedfbbc0e6891df8b0dda39120c3f3a2f802e4a6de2", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/calibrate-serving-simulation.json": "0f526f2ec7308197e2f9e71b6f4f06892f7c22df5b8f75039bc1551c93114afd", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cgroup-resource-controls.json": "72dc255fd021bd2b47a1a70f4640cd9bf2d95670c2d6d9d5666547f707c481b4", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/checkpoint-recovery-consistency.json": "c7cbc6fdc7ea21b9e1fe4ce7fe7cb71a96b0b7f02256337a47fa6080e2572f81", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/choose-parallel-training-plan.json": "4a7a75997eb887d2b06f9856e51f7f12d67e48cecbaee8d115eb6fd69a027e22", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/cite-check-downstream-pr.json": "c88d17cdf3a9e6913b003362b06e4463a331f297eed031bca3ed08d8be36ee7c", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/claim-contract.json": "eb1eeff2f5f6182d90e19e630900da5d6b40b18710fb35e20395a2b3e1d049f2", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/comfy-ui.json": "02518486b39e041578721939a96a819cdbd68b58bf325a61be1b886fe5328924", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/competitor-scan.json": "b2f478677aaf29a9f3888e0e3da869729d3d70b8c421a63efde5d1ea47566dae", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dcp-async-save-performance.json": "5a9060e9c66b48ba16017cdbabbf378a894b5c121ff8e63894030bd29f20db4d", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dense-performance-evidence.json": "96fd8d9530204595ec0114a35a4263d2bef8f40a9c659e8ae2f90fb5640052a5", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diagnose-vllm-admission-pressure.json": "df61666540c7bc88cf45f00aa625af19e88693c5fb035de725920456b3fd06ab", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/diffusers-invocation-state.json": "1fce75dd793c81bf39fb489916aa5956298e8844ec41070098fd8ff705b18a76", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/draft-acceptance-issue.json": "e31e1cce32965be0b617caaf5ebdffa0c63682216d61454dcb5b3b54bd8450b9", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/dwarfstar.json": "ae82754a78426dec87db570865c5ed7bbd60b1c8417905a02e2c961f61f15732", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/evaluate-elastic-serving-placement.json": "7c8d9e4a06175268decbfd128389ac4cdcfd19937f31d96bc16077a808a97f63", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/generation-cache-lifecycle.json": "0e48fed6786a3793038d42d6248c0239de498af5c111211cfa3add4adebb9cb1", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-memory-management.json": "05d926dc59370ee427b52f210fd86d68927759763454bed4b1cc6a6ef949b93c", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/host-overhead-optimization.json": "fc5179b7493fd61cce1203e55e20cf0ec29ea6a8f9a6ce18a43707b1d5ccb02b", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/husky.json": "942e78401db12ca8104e0c0673fe88a5de3226ec846e721ef849c820a853a324", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-benchmark-design.json": "d309913c70f3b025a5c7b531e2091a0729cb64d0e6a3f9f0bb2e7f5aa57be3e4", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/inference-optimization-casebook.json": "a76b0764c0c88fa4840109fb5d4c8e096d43db12296679f3c6e244975d28cbbf", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kernel-writing-guide.json": "95f8fc71c0b0f611eae67a7d655f0f0b8970125bc2c02fde63ed4df2637b67b6", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/kv-tier-management.json": "03f05946c1752b4fe69b36ed29dbfba3f6d0b36fd678be934df7cc2b5ae086de", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lazy-evaluation-stream-lifetimes.json": "061af88c4ce4c7d745e0df0d37c25244dc46079be62fa505ab613d87b7f0c82b", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lily.json": "91492a7f92e139103ab62c4196e1e3b6e523639652c5a91b1ff3cfc55fd90e04", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/lithos-metal.json": "ffe23e31b852af9f1226b999a3ec439cfa6a3eb0a258d36545c759063f267777", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/local-dream.json": "4dda61a92d4e293a5a7a8a7bfa3f466f0a7e18cf02ce4a8e388d17e669d741dc", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/mistral-rs.json": "7d02d7377850c5bce392fa6c476789e3941612804e13ef9bb5095ddd14e687d7", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/mlc-llm.json": "3e16ef715064adace3383a68f7d44d24df9e2f3dadbdc68cd1ef0fa8a9abf787", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/mlx-swift-lm.json": "203ca8af3493b532b431bc9289a0c3fa23261e47e3cc3a648b92508448ab3092", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/npu-skill-factory.json": "3578e2b55013ca096448d660058dd98660ac5160c8b8c2e50433a5ab64105421", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/optimization-search-policy.json": "644f6e9374d5585cd92b1ad85433e42db54aafda82d9cc69939d9b468e9e2b5e", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/os-process-placement.json": "d117d40dcb97be785bd829cb156316c8fd56cc1e352cd2414f677f70eb68cb7e", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/performance-ceiling-analysis.json": "bcedc323625758b207928a08dc3e3ab879bdaa3c6ac6e6ab537cbe47256c35a9", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/plan-activation-residency.json": "f1634088fc0d993e43821ba723710c015bb8e2d189a7c1a1374f743d2ca12b73", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/project-maya.json": "79a1c396629b806b7d1a1bc73f43f07c3ec416bb0cdbe36d3b4fb48465addb59", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/repository-evolution-analysis.json": "4728061d0231c31c7ca176d1b31d145d00c05cb23d298a02d7d76bfaf0aacd90", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/rsi-evidence.json": "4218c3df5a557a5f5075138a82be38ae236447a10e17a7cb0fb7c748ec38a533", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/skill-evolution-playbook.json": "9a14f34e9c7db1f49cfd2c4259345b699490a427dc000a79c6fc5eb4df421a0a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/splash.json": "452829b088734fa59518eaa9940d0d531f1a6da6ff4dd72718468e853eb91d99", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/storage-io-measurement.json": "ec792853f03004c8c1251ae0a07056470796f31db63e9b2e8a6cbfd51d462558", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-input-pipeline.json": "77a07f7077e4de1db461209854fb97a74ca4cc51b4176aa0cd8f9fe45eb564e3", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/training-resource-budget.json": "8d0559cb195ba02bb44e0ec3b274662351830af10ee1ae239b9f99bc3a302b42", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/video-diffusion-optimization.json": "7fe23244a2fa307796ce0cdde9ad3b98f28dc52d6ff9b654b5140cd961ba4839", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/vllm-metal.json": "5f4e67e7404ed606efc155ae3aa3dadabc063d357e87205f96747a32dcf19d10", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/common/skills/weight-loading-optimization.json": "eae1a51b982f360061c0346282481a8fc2c61b25d3808aca3ab442fa2548f7e3", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-afd-p2p-exchanges.json": "692c60c04a453644145c830296187cf10cdbb200f791d92e3a64d808d4eea59a", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/audit-flare-session-lifecycle.json": "a2ef357cd1084423592d7487cf33f5cb72f296733934f52c8e0f555d21344c33", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/diagnose-attention-bottlenecks.json": "6a94ca9f30457847d8dc033b5c52e709f4485d8aca97407420ca723a1fd1c602", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/host-overhead-profiling.json": "5818d8f4cb270e9d1547f19d0abf18de844e8113d18e1b04cd23bf41d279682e", + "src/inferrsi/data/repo_packs/inferrsi/knowledge/cuda/skills/reason-hopper-kernels.json": "f400eb42277af6844f85e9e39b99edeb9ca7f64d4c2f49b2d364cb84578fcee6", + "src/inferrsi/data/repo_packs/sglang/knowledge/README.md": "682e1673fa96c4424f8e1c203e1ef3af174090df4b0f4b00a08458cc835af050", + "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/parallelism-and-extension.json": "1253f5f01bcf1498703c7f34960f0c50580df5ba8f140af557a903e296ad81d3", + "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/runtime-radix-hicache.json": "5a0b2aed51649af214cdc1034f57650b7ce195c934a17452ee67a35131f80cdd", + "src/inferrsi/data/repo_packs/sglang/knowledge/common/llm/tuning-and-validation.json": "ef4d4f7182da4433d59d72dda0d21b4aeafadd73260be11ab8d4cb36c20fafc6", + "src/inferrsi/data/repo_packs/sglang/pack.json": "bce2e0c42796a9ef499d459c37b56d1d56f4b19344841359c534fca29c0359ce", + "src/inferrsi/data/repo_packs/sglang_omni/knowledge/README.md": "7cf067a6162ad39efa42c485c2c41eefaf417883bd7e4039e550c73b28448f7f", + "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/audio/serving-and-streaming-quality.json": "97fc1d5a11a39a0634dae9e9602a4ec47a4d2348e121076c2d5c836cf0692114", + "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/extension-lifecycle-validation.json": "7da8dc5d2655099de2bcd5f0fa922d9e1bcb0bfd2f571f026a004a7f402a75b6", + "src/inferrsi/data/repo_packs/sglang_omni/knowledge/common/omni/pipeline-state-and-transport.json": "0c57692772019504264e39bf0444a69e20cff86f6e112b072c53cf7a4299cc82", + "src/inferrsi/data/repo_packs/sglang_omni/pack.json": "8e70336bc2aac0c74732a026337813652b5d5c6e1fc60f5571e9442345077b96", + "src/inferrsi/data/repo_packs/strata/knowledge/README.md": "b94b6accf0a2333b751f234d7c60c882aa02a25fb9271c5f75905584ccae86ce", + "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/architecture-search-space.json": "2c5ad72d5bab4f7082fc3be1532c347444cde38f5a84639fe16b22134b68685e", + "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/controlled-execution-tuning.json": "6a2bf611a432213919e1c237687507ceb0ab753cccad0e934dc77f86604805e9", + "src/inferrsi/data/repo_packs/strata/knowledge/common/llm/engine-validation-contract.json": "ef89775bc82965fe87a832c7ec8a5cf55cb9eba4ecfc6279caea9ce3b0410c41", + "src/inferrsi/data/repo_packs/strata/pack.json": "643b32defb8db9aa230f624856f4fb4b595b5c23db58d9453091f5420cff71c3", + "src/inferrsi/data/repo_packs/vllm/knowledge/README.md": "25bbde60f83028b9aa8647b8c5dbe30107a6a4531a2463fece99c4a5bb585774", + "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/bounded-serving-measurement.json": "fb00ddac14ef6b06f48da9ba335529bd55d5d83c2a2fe331f0aa1ca16afceb37", + "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/extension-correctness-gates.json": "4ea7083d277df205ea4c361bee87a8207951596e9ef6cfa136cc7d2aeab02f46", + "src/inferrsi/data/repo_packs/vllm/knowledge/common/llm/runtime-and-kv-ownership.json": "e6142c1065754716a5f26f42119722908dc8fff63e667c986d57f1227b8610bd", + "src/inferrsi/data/repo_packs/vllm/pack.json": "1705d96710249c4514c15d8d011b43c1a1ae7c15fbc9ddf11ef2d0e6b2ab7d07", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/README.md": "ab8234fe1d420b072038118fdeabdccdf62a13274d44ed58d79c53db745bd412", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-reference-kv-reuse.json": "80b853fc594d387ab48bed04dca245e0a22b8cdaf7c9b853551ab3bafa302e84", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/minimax-h3-sp-head-buckets.json": "c4d29d1bceb5e2cf86151bf53c9dafb3d83428fb1e9d2a848d852bb8824bf857", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/ascend/diffusion/qwen-image-21-eager-t2i.json": "ef107d43cc448219eda4d64ab961b61cda6350e20d30d94de1ff7c4ffff7e775", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/audio/streaming-serving-measurement.json": "2680f32f28907f6abc7a7e1e3ecab4f46ee412f40f4898ed2f371ffadf47ae4d", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/measurement-and-profiling.json": "c8b25583cd160024c7472a20debea34cb6d4bc79714b4bad2953555adc9f794e", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/production-integration-gates.json": "649b49ff64d3e9e88bd620027dabf9d126ca69bbe91b448f28e845ecbe1f8e2f", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/step-state-and-output-validation.json": "e1ffe28c4fa54af739cdef2fedf4755bba18e078f8380d4fde8138ea712e5778", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/diffusion/upstream-review-gates.json": "5127709bb85ce31008d19468e1b4fcb5454a71776f69f991841cd9ef7558f9c4", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/common/omni/stage-ownership-and-search.json": "e75d6c2829ae22881893f5da2ad7edba099fd64ccfb0fc93d30473e06e9d4d80", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/audio/concurrency-scaling-state-and-overlap.json": "cb47fabb232db81ed525cbb54371becc7e4a28b2bae9d8492f60dfb6f31563e8", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-adaln-schedule-cache.json": "ddbba3bb72b5789938595a82d3280e861f75f6ed54f05a1d2b7dcb1089d30f63", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-cuda-mxfp8.json": "e4c538f03f4ab90a4acc8e3d8721bbd226aec1ee9bde45757a7c27d3d970821d", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-fp32-modulation.json": "abb8d94a1b025f0e2c8e085275e96ee85d68aeb27cb12d0918e4647f8255b693", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-qk-norm-rope.json": "4c1c4acfb9cc0baa90bc28fed1903020ea1ea38755b84315711795db45e48bb6", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-ref2va-teacache.json": "8aeb3f99ac4227cf8b3a5cece13d938b19082e265a557eddae27b7e495bcc22e", + "src/inferrsi/data/repo_packs/vllm_omni/knowledge/cuda/diffusion/minimax-h3-vae-stacked-tiling.json": "8cc18eacbbfd8fc4d3a6b4152c4b584d559fc9b15377da56ef072e542d807c93", + "src/inferrsi/data/repo_packs/vllm_omni/pack.json": "58c971e9dcf6ac15752629ed4d9426aeb5b5d857fc37b5bcf921aa2e81c0cdd5", + "src/inferrsi/data/repo_packs/vllm_rlt/pack.json": "6097041b8d026740060e527f49b1b4e288315b1c3227f7583805b86d8fa415f7", + "src/inferrsi/data/research_profiles/minimax-h3.json": "c5f0fa7871710d80a55120724dbe95544054276e03de65508d6d6e5efdd446b6", + "src/inferrsi/data/skill_library/README.md": "5fab32f90c5f8de6ccde01237de98bdfc7d2b83869df1413a16667d3f6ebb062", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/SKILL.md": "7581b56336d07213a87d01477793ae1a8acf0b7cf876fce36c9a761ee32c1ff4", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/metadata.json": "68ac7f64b7ab947fc2db664bc3990f102c1cc5539b42350aa5818d0bbf17d86c", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/ownership.md": "40688cc44183027efbf2fca8d3548c299cdf2e33fef54f76ab3549353eb9bd1b", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/source-map.md": "10562a75683bd9bd07ec2e37a1cbd0f93f2534d6f85b55a566adb2d97485215f", + "src/inferrsi/data/skill_library/frameworks/diffusers/diffusers-invocation-state/references/validation.md": "5a0b854b87d091d7d505b0bdd5c03330052507a851ab6edf35b93a6675f2d4f7", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/SKILL.md": "a95569bcc0da6e715fd74783a6a1e0489e92b680c582f15dafbf98df589d505a", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/metadata.json": "9c2293c8956508de987c886c7224d673d22912e7e915b026b05abdd287e97401", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/execution-lifetime.md": "c4bc41e08aca68f661325f61e6f0db1760a079dc362f295c847bdc374fee3bf6", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/source-map.md": "798b9cb89ed6b885eb47464ca9f137d2802a54571b9327e631d7742d3e5abd5d", + "src/inferrsi/data/skill_library/frameworks/mlx/lazy-evaluation-stream-lifetimes/references/validation.md": "ebfc00bf7d6ab0a1dc415953ae84d56ccdc713d5614aff85e433c4a843793bff", + "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/SKILL.md": "645139ee0c88ec865cad952e24cda35f4648d4b777b732c9b7d274b6c7830845", + "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/metadata.json": "db7ad368394ccbfdeff6a16e3423c300d51305b1f95ff5c3936df91738a8cc35", + "src/inferrsi/data/skill_library/frameworks/pytorch/dcp-async-save-performance/references/sources.md": "22afe4f4d77f441096cc657d55d763473bfcbea8fb531c8e493fe49ab84955f7", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/SKILL.md": "d494fc7e539cbb7388189bd7b3f04bb5c4983fdd7a82484348aad1cee7f741f2", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/metadata.json": "71c68431663ed2336cccc985a354adab6b13b4cfd6123a18a5d8d53c28268509", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-aclgraph-capture-replay/references/source-notes.md": "6ffbfb68cc7e04923b4fbca2a73a7555b7c3b45a6f40819bb5f3213ef05cf88d", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/SKILL.md": "03829039860ece35519f2e4c339a48b5014ec25e07a2c150a3efb6a3a4e35b4a", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-device-health/metadata.json": "80d80c15514919826fa72d6f6cf26b76db8accd85d21187fc239232ca454e7bf", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/SKILL.md": "4ce9dcadd1ae3c7a8cc633336760d0a3c382f354f1b7028e3a10d28fa0d0c9fa", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/metadata.json": "b6cb25879b1bca5f5af08b48776f2e664cf5dcd11a0efb1c4f56879e2afa9006", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-distributed-parallel/references/awesome-ascend-communication-tests.md": "0cb27a3d24409171bb414d7ebf0c49ad62be5f46482f29624695496ba9eb3e63", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/SKILL.md": "9203cc5a14a1d65df5fbf275734b93be023e06527459b2f79864f28a338cb44a", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-graph-mode/metadata.json": "72b41e7845915d34b09219624e82976386f3125816293540c2c38143fe6b477f", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/SKILL.md": "8fdf46fb2743b67a66bc069d22a45305a222debfb606777bc451b4b8f31d9cc6", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-model-adaptation/metadata.json": "be681a3206e5e183b4c9670c9d7784bcffce18c891449e5b15f94ca5f1324b9a", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/SKILL.md": "401981cf7828edcab0c9ea882f7ecd83badfa3c2a95006e912cf99d67b3d6a12", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-profiling/metadata.json": "efd37bb23acb500dad4ec8ebe7ddfe5c3f196374052825f923ac85da8445f06b", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/SKILL.md": "38b151bc6b4e9734c6fa41efb7a30343559400c0c75bbdd26bd8a968b8ea4592", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-operator-tuning/metadata.json": "04d2002fc8a0a7becf86b0b73daaf2da8716ab466d7f988da43d41a10f26680e", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/SKILL.md": "647f5e9a288f2c4e1617ba53df59d9d827df749eb4fe86182d21a9c41175caf8", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-optimization-casebook/metadata.json": "f425f7c1abbd616387273e8932c6bd02f4b1da012f099286ad4171e0f6037b22", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/SKILL.md": "a057b7f86a24f34a8f84abbbb4910d58d4e50ff47060e1c404600017e3febe89", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/metadata.json": "9e705625c8308013c287576b0247da6f6116d817a4a535cbecff80abd6ea35ee", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-CANN-2.0-LICENSE.md": "89b1c114df62cfd183004f38d18d07f186436c3b37203dd69c3e84becf9d2282", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-performance-analysis/references/cannbot-stage-evidence.md": "f56e4125325c9c1b9235a1d64fd732861c7686444492422daff44ba8dd648be4", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-video-diffusion-optimization/SKILL.md": "53c21e64f0dd47f9510faef1f2c01e1283899787ba7ebfe70cdabd8e57a76f00", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-video-diffusion-optimization/metadata.json": "8c2b364298330751fd1252516159b16c27e90354e2a724782d2851e68e43cdb3", + "src/inferrsi/data/skill_library/hardware/ascend/ascend-video-diffusion-optimization/references/field-notes-video-diffusion.md": "dadac7b23462fdc325218c72ddf8a2ac553151761a5d3e45b854b53d2c8c6908", + "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/SKILL.md": "93aa1b775ced304cebfe6c4dec0503ddb2260ecb3cb01edc17ce329446c5ce2b", + "src/inferrsi/data/skill_library/hardware/ascend/ascendc-tiling-design/metadata.json": "f6d94f104cb70969024b2f4542078a240e200712c66279eca8b94f60bee1dd71", + "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/SKILL.md": "0e6f094dc554abea12e07def50f20b233b47127779a8aefad595b4ab1efd24c2", + "src/inferrsi/data/skill_library/hardware/ascend/cann-attribution-overlap/metadata.json": "acccc8eb86de0df6caf78a9b3c6db76a2a686adc4a5ba0029a4d62d50ff9a8f8", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/SKILL.md": "8428df7157af024fa17cfe0b3818ed54bd9f468306dda46eb8d4f0edf4f3f230", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/metadata.json": "c650dce88c07b636d68e063dbda4d319f6f7f3ecbcdd5e47f393cbfc3af7c71d", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-device-memory-lifecycle/references/source-notes.md": "80cba4d3a884e7ee2618326cf74b9f9c3a34e4d9cf443707f6bbe673f2d5ba65", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/SKILL.md": "23e93ea7c98f99743509fdc04139d78ed2e7123c63e80b366c6beaba7eb70d0d", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/metadata.json": "97836a38465106d816a9ace894a7b277d6d330f9accade8f8493ce6f0fb57ef9", + "src/inferrsi/data/skill_library/hardware/ascend/torchnpu-pinned-host-memory-lifecycle/references/source-notes.md": "435193590b4a2ba32e2e4c6fa9d7e8f889528d412cbbb48e35ec5433f7dab384", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/SKILL.md": "a25738dbb3b9daa15c5827cb536a20fe13f7c8b9dd14aa3fb7cfa7e132879431", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/metadata.json": "3abe626a6e4c73c943429f609ce2c62aa77aae7e1726475ebbbdfbe1794f6f7d", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/mechanisms.md": "fe08b2292fa587be3874774ddf41063b84c579a0085df422ec6710d8aa90b8b7", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/INTEGRATION_NOTES.md": "371b01ee201ef51fecbfd6da892d517aaba295930b9395025f03f34e21c030a1", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/NEXT_GATE_LEADS.md": "35b8b58ae5db02b2df51ccbe5cdf870a7383d17f345f9ee5eb7b0a4146dfc74b", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/README.md": "e0b1ec1b39c10067f5ddce58320851ad089568c4b2296a24777389097cb5b388", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/io-reduction-can-plateau.md": "5cc35f514224ac6e6efd2da6db221bba468b2f4db66807295e67b240bcdfdba5", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/overlap-and-fp8-counter-signals.md": "80c8a60a6df55f3d84fab23e03567d8b38e23bcdbd63cc09881e46db1e2888ab", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/cases/partitioning-after-fusion.md": "4d4cabe0991eab554dcfbd4fa9c8199e61fa4767709b0c4cd646e2588bb398a2", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/papers-20261015/knowledge/attention-cost-model-boundaries.md": "5ed8f888443e38ccb72545b6025447aab7b92936f78c151fed680e582126239b", + "src/inferrsi/data/skill_library/hardware/cuda/diagnose-attention-bottlenecks/references/sources.md": "53436e0de4e6dda2d29426591bf61f8526378cdc6f9ec7967b88b1bcd4c61e6f", + "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/SKILL.md": "326e566fbe30d3ba04e4b50c3e87aad9df037f0ac4ad6fe129165ae97dc5f91d", + "src/inferrsi/data/skill_library/hardware/cuda/host-overhead-profiling/metadata.json": "76e34c1c64aba473a9e126b032a776417caa0f913f6ac3cf4dba3a5d123b1ee0", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/metadata.json": "7520da1cc97412631a600957ecb18aa5a7a4461ea21e76c237330180193ed898", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/fp8.md": "31d9c211c11c2224007abbdcd4059913c4a54833fe2d2f926411c5238fcd3501", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/cutlass-bsd.md": "42fec630f410aa308f70a51a89fadcd19586fa620f9831a32bee528a9a10000e", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/deepgemm-mit.md": "6c73a30c14a2938abaa6afe98ebb19e6d81dac5bd4ace0cb05c61c20afa37b85", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/licenses/kernelwiki-kda-mit.md": "84aab57158a914f448b38d8be29dd842b634a08c192d844f416a256beed4ae94", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/notices.md": "876ec016644bf5331b3f257985e126c0e61980e3bfd82569b7b0c3f6b2c6c64c", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/resources-and-scheduling.md": "1ce25b426736072435a1ead9a2a8e1f56df5d0abb8198e8c4fdcb34321db4964", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/source-inventory.json": "1dc09c895c970a0c9708158bcc6dd65cdb736274fd2c35d44542c70407d3849d", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/sources.md": "6b74b9e774f2a80db0860510d9e6940d36284430f29c1e124f115a9eaf25833d", + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/references/synchronization.md": "4322f6b3c0e4368bf8c192b4ec752529767f50d19c0c38dad8fdf9cd781e41d9", + "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/SKILL.md": "fd1f5a7e6a55714730044155f5b38c1d700383d4e16c3e91963bbfbe1c773af5", + "src/inferrsi/data/skill_library/hardware/host/cgroup-resource-controls/metadata.json": "19e40763fedc20990651bb989c32a7693306c5c54af46c3d4be77fd4f87b6c03", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/SKILL.md": "2c807f3df2da2f533d25183de5771f3d412058d16410994292ef8f48c49f9b3a", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/metadata.json": "9ac6af135f0b6d76a76bc201b6bbd8ac9afe5201dd1e2ea82c0ac4826de401da", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-LICENSE.md": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload-readset.json": "cc6c2011be746002ba3ff96ccefab1809863748ddd17eb102b75e7b0e134fe55", + "src/inferrsi/data/skill_library/hardware/host/host-memory-management/references/vllm-ascend-native-offload.md": "6c2a72b5ba3708a01d229e0576570837a617e6a42dbb0b5312a26a65a6c31542", + "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/SKILL.md": "36ef88826a32ac3147b869001004ded17a8d5a491e25bfe105005ba39ea9d1eb", + "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/metadata.json": "ae71c389dfdb1a19d267feecb4c6f1fbb4c9afad501ff6d8fab7e38b29bcc37a", + "src/inferrsi/data/skill_library/hardware/host/host-overhead-optimization/references/deepinfer-zero-overhead-scheduling.md": "f452bf40132c9195d739f1c86b3aa18857138c7ed0f1861af5853d93b9955924", + "src/inferrsi/data/skill_library/hardware/host/os-process-placement/SKILL.md": "2003492e4d89fde0443b180d74933e764bd1979435e109c8d26e20eb8f6e5053", + "src/inferrsi/data/skill_library/hardware/host/os-process-placement/metadata.json": "ec02995c075f0a540921f57b4846f2cafe334eb90f44389f17c3c50457c3e9c5", + "src/inferrsi/data/skill_library/hardware/host/os-process-placement/references/llama-numa-thread-readset.md": "748b254b3b043a5ddb0f480eb4a566854d27eedd7b111499863e9bf0a5a6951e", + "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/SKILL.md": "63b883c10a8d3e7d918abfc7175b72e2b401ff1e02697f94b8382ec1c8f369b7", + "src/inferrsi/data/skill_library/hardware/host/storage-io-measurement/metadata.json": "0a41c7b00978332b54ee1e4a7c1ade88ecde3e961128f24d1b4f7708caff5ff5", + "src/inferrsi/data/skill_library/methods/assess-serving-preemption/SKILL.md": "0a5f48d9352c5bf460775ef47f9136829978811f4a16e84983e299782cae305d", + "src/inferrsi/data/skill_library/methods/assess-serving-preemption/metadata.json": "77ce5ba16245cbd80d46d0abdf80aaf26fc0924fa2d2a5624ff430de68a1daf0", + "src/inferrsi/data/skill_library/methods/assess-serving-preemption/references/source-notes.md": "f29fcecfba0490008a354c11b42403983fe165e56ff907ad61cc6040d2b6407f", + "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3", + "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/metadata.json": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7", + "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/references/source-notes.md": "61b1609f71c37b61d136b8c4d39f793de87cffb1cf31c8ec737a1c3fa6fb1b4e", + "src/inferrsi/data/skill_library/methods/audit-fixed-workload-claim/SKILL.md": "2be89b1340a3deb978c8e1792d18aeecbe8f5de81d4daf542389f53826b8ebee", + "src/inferrsi/data/skill_library/methods/audit-fixed-workload-claim/metadata.json": "81edd75fa16939807548d0944e9203cfd7c0cb17e63463872d96ca2c1273d1d8", + "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352", + "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/metadata.json": "261ce9f644ea6eebb90c830f41fb8b320d2b401456e5bd8d494095a44b8620d0", + "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/references/source-notes.md": "9696d3c0ef96b6611787446bd6f96d4d6255ca8d107105de8c1d9af83edb9fa5", + "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/SKILL.md": "6b7db1eb2780dec27858cb7aa0a3c093c5b5dfa4c30261aebfeaa255213ff688", + "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/metadata.json": "4dc8659806539b4784ea9348a01d4194ac70d0bb297a916f90ff6419efa5c328", + "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/dlrover-persistence-replicas.md": "055f4357f79026fcbcc766bbe77db610376ec353623f8898f5d14a30ee69e7bc", + "src/inferrsi/data/skill_library/methods/checkpoint-recovery-consistency/references/sources.md": "9425b1aab963a2c8d35f3db897bd15ee69699bf5b0d0c8425c3c8bba3e9a0233", + "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24", + "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/metadata.json": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7", + "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/references/source-notes.md": "9f9184d46b19312138a6e805e439e43dc0594a0cc6b46675255a582dc946ca5b", + "src/inferrsi/data/skill_library/methods/dense-performance-evidence/SKILL.md": "53cc1867a13f54d1006d50517dea8848869ca50df3f0cd610b2424cc8d18c933", + "src/inferrsi/data/skill_library/methods/dense-performance-evidence/metadata.json": "a0d1e3b17140295510a764045d1b222e19be3a3406e6d1e7dddaad49ff474490", + "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/SKILL.md": "82d0099d3c33706f46c657d4fa7f76d3d486e019b36760d1de001921328a60cc", + "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/metadata.json": "37379a4d4a058ca46da2b20a920dc01be9ec11e557c32b2332702ec76027f4b5", + "src/inferrsi/data/skill_library/methods/evaluate-elastic-serving-placement/references/source-notes.md": "ea416831fd7d0874e06d0a76fb0c7196df53072b717673a0cdf635db6bfd917b", + "src/inferrsi/data/skill_library/methods/inference-benchmark-design/SKILL.md": "59b5656e7d989b3cee8d6f95cd43be03053b1fb11c3a82bb49ab3d402f7aacee", + "src/inferrsi/data/skill_library/methods/inference-benchmark-design/metadata.json": "18310402a8ade03fd48ac687f79ba9d447871e90337de84c0e1bc15c0b827306", + "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/benchmark-commands.md": "0898424e0bff3a6246e1eeeda6abd6c5bd974c047fe32f3bbd5ab9d31ce17cd3", + "src/inferrsi/data/skill_library/methods/inference-benchmark-design/references/task-benchmark-selection.md": "0d4bb080a754e26846e81680023db0dff5e6a4c96be4327c411c3b72ac4ef8e2", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/SKILL.md": "89a6a82c692f7f5192f4d7f4c3a5ff08a4df3b1c4bdaa0aafb7e04292d8aa5b4", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.json": "3319e23bbe2bd40e2d45fb89dc87bdc92d4df2b5ddd8086fcf84813b43885b7c", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/admission-drain-occupancy.md": "e88bbac5d253a8e7e3a3d5e21047f23f2970c11ac23887d4303ab94fca52dd27", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.json": "2cc0a526fe0509551607a03e7626dd3d7844e643c4afb194837f75cddead8b12", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/drama3-rejected-directions.md": "6f143bc363a3290fc008c08e57cb075f1776d1c8cca8a6a59297d5b1fbbf61bf", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.json": "b3d538f79c56724ccb2895cbf50cf2e6b0fec3979be62a684a8230cd932dff55", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-ascend-rejected-directions.md": "be396983812262515899a3e2ba4f01b23ff92e1b836881882c7580372b8ae552", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.json": "146ea1564a46ed03092c7dc5796b318fc05b04b1821105ee00eb7fa7fa1333e2", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/cases/qwen-b300-rejected-directions.md": "cce2930f544a0f78f21797fba4d8ded298e4a81f76c34d7977b0e31d361bc738", + "src/inferrsi/data/skill_library/methods/inference-optimization-casebook/metadata.json": "fb55067fb45970319787068a3206110641cceaf0ccbabd706be6d0b3d2afc2ed", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/SKILL.md": "360361f592ec3a24834198372cbf92aaedc1720404d182df94315eb0e01d58dd", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/causal-conv1d-dispatch-state.md": "2eb67dd522021bb2234616d7f660e6ba99acb43676f48cb23141bc69bf3e0029", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/hf-qwen35-gdn-binding.md": "b3a55af3f45f01c48e1cb18c8b029e5949a494c3e51dc444a5cccbad9acbba0a", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/sglang-gdn-source-contract.json": "6c2586ec80698dba848ed09024954d060307be71c8701719af19df3b735cf595", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/sglang-gdn-source-contract.md": "8b093faf0ab775d5605ecdd9653fb3689c2b839cba2641c5d5cda17c721bb2ac", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/vllm-gdn-causal-conv-contract.md": "2f3dcee66e2dfe7a98bee21a36b892538998de5de3caf1bc63257137f9984236", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/vllm-gdn-source-contract.json": "73f1a2d8fa89ce05aead2091a05a285b13cbdc314d05452c35c6a2ad15ffeff7", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/cases/vllm-gdn-source-contract.md": "7f9ebe8de7e55d3db2a119b6a56e15c37448fd7d4e022a47122a143206d31687", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/metadata.json": "c86f93f4eacde789e5438327d85b1de0bfe623fb87acad49306e47b7e8ef1261", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/akg-affine-tail-guards.md": "113deb2db10b171db3425bb3f1792bee763706b794c8cc2e222b4cc0d0114118", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/causal-conv1d-contract.md": "b3a6245cf5e9deeba0dffd4a15ddcf97da6a4315f4d3131921779438832663ad", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/hf-qwen35-gdn-binding.md": "848d3c99b47ef711e8820bbc4f6bcbd1db7211197811c9d855201025448d806b", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/hf-qwen35-gdn-source.json": "df3f4650b5cf92d4fd3d5a4cc8aefb91dd4b0c2affaa0c873138d5651c50e6bd", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/licenses/sglang-LICENSE.md": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/pplx-gdn-caller-context.json": "0219fc9ad785da4e638c1680bc9eaf28672aa32313824f52bdcaed61d29bca87", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/sglang-gdn-causal-conv-contract.md": "c9f27b0231ddec3d14d6fa0831d1494ffd8e11bd8ad704e04be3bc4a4ffc0e58", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/sglang-gdn-source-files.json": "4c7ec2cfd2c29297e185ce09b4d9b2040fbe3635943c112f9ebb71dca8ab381b", + "src/inferrsi/data/skill_library/methods/kernel-writing-guide/references/vllm-gdn-causal-conv-contract.md": "eec09a6e95983c46b67d5f066d34419c17819899f5d8e3c65f2f163f2ada0fde", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/SKILL.md": "dd275a0576e827353963aae7f0b09494d60d0b9a4a2884fe575c85249fa3131b", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/metadata.json": "4a61aab6b9984dacdccd657c17f2d959b3a6ffe74bfd2f1677438f3bb8f025d3", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/INTEGRATION_NOTES.md": "157e74ba180426b279b548a729769f152bf18f9a78ffca4ebfcaf26b7d0c5292", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/MECHANISMS.md": "18bb2f5d7e52773c516420a142edfb531d0b9c822a2934d60f3234ac1cb9d40e", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/NEXT_GATE_LEADS.md": "16a264c7159f9c2c700ce8ba46e63f3e636b277c987da4970f6d2fa085fcb053", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/knowledge/adaptation-interfaces.md": "5e761957cb5ecd93926e6ffc13deee7c830c6e8f0ee3c3846e9dd3b50dc74446", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/references/source-notes.md": "54d83c3713045d59a48821bcf93598374ba4a630e3891ad39498f919317c4fb8", + "src/inferrsi/data/skill_library/methods/optimization-search-policy/references/papers-20261015/updates/optimization-search-policy.md": "d3db44fa9a6831b86a02c638b323ed943e69fe256866213f10d68c72d762879e", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/SKILL.md": "c5f78961c23e3c8d659f5675b522518ab06e1f2a8ea8bc74c5d428ffb41693d5", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/metadata.json": "60b5b24ead590b0fb222398fba47822bf6723690da683e256a450bb5c0eeb806", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/INTEGRATION_NOTES.md": "083e462e0a204dce599362d46712fcfb392d053c5cda56216e22e28d4b8fa0fa", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/MECHANISMS.md": "742c249082ceca230326f4407a40130fe624553b796ae5a157a8a4c15a7e1398", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/README.md": "17b2d1c7c14124d59ac8a0d5ad4200ad2803d991e784a1a98cd97188cc364999", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/READSET.md": "66459ce393858b2b2160f1e2973373bfc4b9902a12644631934b4ebe3140296a", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/STATIC_CHECKS.md": "cb0aa7d148352ebb7b1678fc42a557a099575054716a341dfa25b6a998029064", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/cases/decision-cases.md": "781a675a6533e0a18c044aaaee0d010da7013b6504654b8b8292a35176db801e", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/procedure.md": "ce036a4d8a0908da204515c47e2ff9623548d09de264fa65bd58fedce52a78a0", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/govern-optimization-experiments/references/source-notes.md": "f9a8254e45473ddb2b73e576957b6d3db83c62b99dff1922f4d1e606e569559e", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/knowledge/bounds-predictions-and-decisions.md": "5780d6eb5257612ec45deb94f85f334a6ecabf7e466d1678706041f6549bda79", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/procedure.md": "6593a63fce9c474b515663a560c94b7df048dc5f6e77a95b8e54e4a64cd10a09", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/redirect-cross-layer-optimization/references/source-notes.md": "6ff3a7bc31f1172db92aedaf7df1f1aac688bf769227b583a4cd82bfad52da67", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/agentic-kernel.md": "7190573b85f94a4391bb0d887ccbd37a31b3f08bb0531114ad7afd72fe1567a7", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/flashvector.md": "3926369acd88ff3c370bee5718e74dd15f3219c3e4f6311095e30e917bffa4cf", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/inferencebench.md": "21f47111197a0b7f8d6384ea918d139dc8318a62ca3313ce060048271f50b639", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/prior-coverage.md": "35d47ac078591221fcb666219bcd824f9400fc17aafecdc0499d8336f7178238", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/papers-20261015/references/vidur.md": "d38602af5219f5feeb49e6cd6bf53d66098c77fad141dae676f1d8a6743cf25d", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-contract.md": "419ec4bcaec45ac579a133bce083d4f48779f26e664d7a86f720fa65f343b1d1", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/references/roofline-example.json": "597cd3205df043cff3ae9d15924c77934e8d8eb300c00c4840f6bb2921f90faf", + "src/inferrsi/data/skill_library/methods/performance-ceiling-analysis/scripts/roofline.py": "524de48a6764401695fe1444f1fa3d38f363069b0cd2fba4715a86ca4b28a5d4", + "src/inferrsi/data/skill_library/methods/plan-activation-residency/SKILL.md": "542d8a1e6963f5019986e1fbd79d3a247a045a51487c09313026087ac457c15e", + "src/inferrsi/data/skill_library/methods/plan-activation-residency/metadata.json": "c6acae57d70190cea0e18c71c79127962181e58169e011df75d8635b3db4daa7", + "src/inferrsi/data/skill_library/methods/plan-activation-residency/references/source-notes.md": "5ff52655cfd06fcf970ed2a7d4d4fe1c31115b102af2d41497de0c74848ed2b3", + "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/SKILL.md": "61bc684d67d6c0876c4956057f11c88802cfc4b771553d00411634d8ea226b3d", + "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/metadata.json": "98ef753a0e74c46f59a69ae267bdc0a5d293761e6f27edc1822fa8bd4548c4d6", + "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/capability-presence-audit.md": "c21eb4a6aabe811a49a6540a34c54dbcfdcf7efae3052a79b54034c7838026a5", + "src/inferrsi/data/skill_library/methods/repository-evolution-analysis/references/rtk-consumer-path-transition.md": "a70b1d128fb5e0ac12292a290d624f218a0831d7581fe8980b62661473e0f2b7", + "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3", + "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/metadata.json": "92b36338b9257e2e41b95e4370694d7de52ed84bc7971056b3bfced07b0993d8", + "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e", + "src/inferrsi/data/skill_library/methods/training-input-pipeline/metadata.json": "80659099fce6a6ec05e9084b75452c877a5812e30f60a78f31ef8c7efcdc1bac", + "src/inferrsi/data/skill_library/methods/training-input-pipeline/references/sources.md": "60e0f3e88655d8e1d5eb4797f39ca4370deda01614f265b8d2ea4559d8ac020f", + "src/inferrsi/data/skill_library/methods/training-resource-budget/SKILL.md": "bdfa5a4ef37d9f303c3a23cd6f98c21137b97ac5ed58e0ed5fac6d71d33e52ef", + "src/inferrsi/data/skill_library/methods/training-resource-budget/metadata.json": "d42b614af9cee6ee0d4a25dc5d57c7b98438feed1e807a7c9f1755109a86e9cf", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/mindspeed-historical.md": "eb97f935ed30517a8fc9edec5e9cddc04f5d679d156c16d75bd29c534593360a", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/INTEGRATION_NOTES.md": "c55c6ad2131ef788866411dcc7056f221420ebff30c95c6bc919b09a4818cb73", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/cases/paper-grounded-cases.md": "19bd25b1a944adb2351ebc6b1d795150ea033041a44f6cc261e881677812a101", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/knowledge/offload-delta.md": "f8740fc5c8bbbe5ac13ff1cfec7ca499569d0b46e8cd56cf1f5a44c3a5e72242", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/notes/training-memory-scheduling.md": "9aa6d3e01452dff95c8a8940ff62ead8c146ce215c9175b78c9f2ef508a7b2de", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/papers-20261015/references/source-readset.md": "2a5c166767f56fa47e7eba7386f1f36da8cb73702e5b5bbe14d6a456ca9ebfae", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/sources.md": "6d248f58af8b350fc66c22898988d6f451ac60d03bcd805cd6e84df5e55174ac", + "src/inferrsi/data/skill_library/methods/training-resource-budget/references/stas-decision-notes.md": "f1a993e6c4ca9a035a2effcaf5500081f227b0cc56cc885a6b22fda0041fc744", + "src/inferrsi/data/skill_library/methods/video-diffusion-optimization/SKILL.md": "fc0e7baf9913ce7448247772a1e224a9cbd827492cf9459b1f17f492b9f67b70", + "src/inferrsi/data/skill_library/methods/video-diffusion-optimization/assets/campaign.template.json": "1a37f1d136612143ddab8f0cbbf424c80433b4da5312edb18d07c0b9d87ef657", + "src/inferrsi/data/skill_library/methods/video-diffusion-optimization/metadata.json": "b27d69142dfea5dd2820170475c2bf6cced2d5eefd1c43b3413c41849af0fd02", + "src/inferrsi/data/skill_library/methods/video-diffusion-optimization/scripts/evaluate_target.py": "9e84df051e03d567746fd73ae7243efb2f3c5ec806fef824d3c4d42a8eb9704c", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/SKILL.md": "a1bf0e2ec066d5f30225c6cc5fb603f830f48eb5e2aee3fc3d084ed9d8aebe2f", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/metadata.json": "6df47ea943e42a11807c54869774957ac2383266120caad730f788207403581c", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/fastsafetensors-transfer-readset.md": "5a39f14b102b0cee304541e011aa31e44e6b13527d1bee9987c3f3b66c064d3d", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/flare-preparation-boundary.md": "1d7922785a196bdbafdc9791fe79428465a5acc0292ca8f99446f8c764b35454", + "src/inferrsi/data/skill_library/methods/weight-loading-optimization/references/llama-loading-readset.md": "6c640f28ee3f0253a17b81461c865cb9f947367002cb75bf72550c4572d02e97", + "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/SKILL.md": "a46110891eca860915e314d6921a22856d631453b22b17837fe30262d6780535", + "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/metadata.json": "4e6100dfb9fc7bc00b19915f4fd38ad509704f2a609c658adf4d632d4e05b6c1", + "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md": "0e8566ea362276878e1f38f97da49964dbea301c15d4aa9d727f97a4c39fd943", + "src/inferrsi/data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/source-notes.md": "5073d87450cb60d1b245308dff8807e73749c07de976d50f2fa26694fad81a72", + "src/inferrsi/data/skill_library/repositories/comfy-ui/LICENSE": "3972dc9744f6499f0f9b2dbf76696f2ae7ad8af9b23dde66d6af86c9dfb36986", + "src/inferrsi/data/skill_library/repositories/comfy-ui/SKILL.md": "c993bd6ff8e9b5f82f39f175ec8242341136e4ee0808af414e1d94ad3e196c5e", + "src/inferrsi/data/skill_library/repositories/comfy-ui/metadata.json": "74742851a002a51c208362cdcd2a59fdab5d1734d8492b1f93fbec3cf21e0841", + "src/inferrsi/data/skill_library/repositories/comfy-ui/references/graph-cache-invalidation.md": "614c9a4e6943d24d7d7be7d632682ce8b1dcc239124a37d9dc1d487113772eac", + "src/inferrsi/data/skill_library/repositories/comfy-ui/references/model-residency.md": "014c68386e7cdfdb7b4f566836e8b2ff61bd5a5a506e7b4723f83fe80c5d4de7", + "src/inferrsi/data/skill_library/repositories/comfy-ui/references/repo-provenance.md": "ca996ef515af42a3ce304747fa3f12a31c00533deaace5ba2ce21aefa0a62e2e", + "src/inferrsi/data/skill_library/repositories/comfy-ui/references/validation-boundaries.md": "1913dedbeb558d872d6a0e015cc11d822910c3cb259a236db29fb60ad8a64263", + "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/SKILL.md": "4703beb6e5194ef2ff2c0b5ec69f64e35417a4bc0dfb50a10c80a920b77ef3b5", + "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/metadata.json": "014be24d0e669f4acbc0717dc323eec784026cf185186619d5a19d6d84496131", + "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/source-notes.md": "fd917bd5ac80bee9c9f4525f8e0300f6d67dc144e2a16632af42a67f3d53d098", + "src/inferrsi/data/skill_library/repositories/deepinfer/bound-windowed-attention-replay/references/validation.md": "60e71e70c5519bcdf6c171eec2be2f342e8801bedd746d6005f11adc646dfaff", + "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/SKILL.md": "f157562d1edc905cb4fc07246567e3c5b844efb9fbf5fe39f0d40553a1b1a307", + "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/metadata.json": "4d68bf4c1767eaee12ade963bbd551d0e61d1bcd81a984bb02fd3aae46965919", + "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/source-notes.md": "6fa955022f16cd598f44fd6521901d1cb868b06cec545cabe2b82769a641b5d6", + "src/inferrsi/data/skill_library/repositories/deepinfer/budget-speculative-verification/references/validation.md": "673d93f6700052974b340fab25b3438e0e6344960569ac9abec7d7720f662288", + "src/inferrsi/data/skill_library/repositories/dwarfstar/LICENSE": "4a8a1358f5b048627d71e4988de0bafd386f4988991929629e6cb9e29b7b9b2f", + "src/inferrsi/data/skill_library/repositories/dwarfstar/SKILL.md": "a04a252ecdcd4710a3273595f4bd1f3eb8e24d5bb1bfe005f058a78388254dc9", + "src/inferrsi/data/skill_library/repositories/dwarfstar/metadata.json": "e1f0c2dc3d7ea9278039c2e8c3e8ca4f8e3aba33504b7784d628e903f0b5fc8c", + "src/inferrsi/data/skill_library/repositories/dwarfstar/references/repo-provenance.md": "7d72f9c85ee60fe003f5f5a585b52e9989ebcedca625b584fa39a976d4433998", + "src/inferrsi/data/skill_library/repositories/dwarfstar/references/repo-routing-metadata.json": "8e195b420b67151932af5f435c284f49677cc7863e650c83cfa9e762fd00db0b", + "src/inferrsi/data/skill_library/repositories/dwarfstar/references/source-identity.json": "a38fe07d003accd83b7626d7e189f7b49b4b73e7a223e4d441b0e33add5d9c91", + "src/inferrsi/data/skill_library/repositories/dwarfstar/scripts/request_template.py": "62e4cde002f2aebeaddd63e6e1f04fc727c271d2cbdaa00856158baadee5340b", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/qualification/SKILL.md": "3b8eb25f6b8e4729b48cce5619daa24422bee251f119ea7f9cc1faf0097ec6bb", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/qualification/references/troubleshooting.md": "cd832ab2a420490524251a042c0c83c2f75bbfdc12e914b91da8f91edf6860e9", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/qualification/references/verification-contract.md": "2801f6683c2974d1cfa5b86c6268fd33f175f1cbcc9ed85c932e9b3504fac3e1", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/runtime-memory/SKILL.md": "ad1d34d429baaa41fb72c94ff4e554b26b2f5b701cc4deed03f1d066c7d3fc52", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/runtime-memory/references/memory-ledger.md": "765915ecd3a318cc7908658d5a79bbdff615fc24409501febcb7ff8af65a1050", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/runtime-memory/references/troubleshooting.md": "42e4785b7d3c1375efa324583897327f28ef76c0b7788092da92095bddeb1440", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/serving-state/SKILL.md": "71c27212847146376d72e6cc6d22abe2f7eed56402c5917cd7dfcda0b2aa5982", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/serving-state/references/state-contracts.md": "8f74a07c947021e3069468268f6b5aa2548a17db518c5c0505b84dde73a1f31c", + "src/inferrsi/data/skill_library/repositories/dwarfstar/sub-skills/serving-state/references/troubleshooting.md": "c9eebe4109d418c4dfb9bb2425c145ebc1a52da9b665e0556149bc31a442e2fc", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/SKILL.md": "02a5f7c18e44201215ecec17c6ac96b27be9148bcc189df287f39d1b372b7721", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/metadata.json": "58716737446f46bd1de77aa323c5652f97277d44dd952c82b42399b2cb769b87", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/references/diagnostics.md": "0908a0eae53a2e10775a8b6ee3e3944b2fa410174402b5a1a8c98eb674dd4b4d", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/references/lifecycle.md": "107ebe3c22e5dba8007c3b8ea61149f10bc55c51b557a4230f3e5de859bb0f7c", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/references/source-ledger.json": "61e51a97ffd7c1fa7866bdd440b024d0c5e0065635ac8d0ba53d2c69c7f9e470", + "src/inferrsi/data/skill_library/repositories/flare-local-inference/audit-flare-session-lifecycle/references/source-notes.md": "fc7ceb819b91ef1731d8b7783f543c92ef6bb5c97ef5f897c055824f8a560168", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/LICENSE": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/SKILL.md": "fdb4d9483c6efa836f9e59617da605fa9c2186e11c2041b05d5c81850ba0475d", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/metadata.json": "fe626a77af3f084f36fd73ffdb6f67976e7dceeafb94ba22078ef03c3a96cfae", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-apple.md": "57ce01754ba96c378f13a91305ef3d418b926f99b2753f0af99315572bce063a", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/minimax-h3-weight-inventory.json": "fbaacaeec340d3ce893556ee864405a0746af4d13530a38cb3d5aa9d60929271", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/phase-residency.md": "2836a550eb2be7f5a862f51ec7b135ebd9715bf3c4faa84a7ecbc9d9c11f5d8d", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/precision-and-streaming.md": "7e166c89193f4c7aed852ec95346a2de2b5008dac002e2e1862be1b3e57f6093", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-identity.json": "730f8dd23648a3f8f47b9726cc613d893a79d1817d515eec609a9b06d8a9ff90", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/source-notes.md": "7304bf9369efc157dc7aa40686dd828cc14db45c49442e908b3d3cc4b329eb86", + "src/inferrsi/data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/validation.md": "0754873a758e048480938314894e25b504ae0c269549312e7f8e6020d83fe294", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/LICENSE": "663e51edcd9a3a8f858018ce6924ad2895a24c269c26c39ad75325bcd0f7e5bc", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/SKILL.md": "e5205de6eba52b3d1693666af755640697717e52ee7834e1ae30e4071b208d54", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/metadata.json": "86a67bb56110c3a9e191e0ff576194162e960cbddca0a99c593ea7aef8df8a4e", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/THIRD_PARTY_NOTICES.md": "a0d77db7d2aca0eda811ab3c386cc32b3143f2c2609a042cf4642954f0f6fb09", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/preflight.md": "5cf745005914c74b18a09f127d1d02ce0cb64f6cd7899501ce3cb40938c42a4c", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/repo-provenance.md": "6a4c14207223dcc2584f7133bd14ff4590c2258fc2bd44f3721b6bf63e3e35cf", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/semantic-contracts.md": "3f518160e8a227272e698486e6c67acd364af0d74755103ffc868b2d66052c20", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/source-identity.json": "b0e63197725c6a1baa4888ca3c48ccd4926b6e6f0f8bf781ae76ac97ffd1ceb1", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/source-notes.md": "f9430103644eacd9c55ace6f84d8587fc0bc063bfcd9cb60e6045b139a9fee05", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/references/troubleshooting.md": "ed552d87f9fb9fb2c56de64df61f3fc6e2333cb049116cd0736f784c8a327612", + "src/inferrsi/data/skill_library/repositories/h3-metal/audit-h3-metal-fasth3-semantics/scripts/audit_fasth3.py": "05bfb9348cf88a76f4fba8bc8a600fe54246d83a98c1685bc4847ace28e2fbac", + "src/inferrsi/data/skill_library/repositories/husky/SKILL.md": "ac86afdabf9adffb684cb635971934438724bdfde8628ec549727ea23c40cf3b", + "src/inferrsi/data/skill_library/repositories/husky/metadata.json": "9b720bc6fcb65b6dbddba36d6187b4bee86346a1bc633f55a75bd43dbe4fa04e", + "src/inferrsi/data/skill_library/repositories/husky/references/author-report.md": "4c1351ca42b2535a7c03c4da87706040650ef10d86eadcdb496e881a147fc0f7", + "src/inferrsi/data/skill_library/repositories/husky/references/repo-provenance.md": "37de912e2e2a36b44233560e6a8368b994428bff587ed4f04b1cb5fbaba0a9e7", + "src/inferrsi/data/skill_library/repositories/husky/references/source-availability.json": "d3523e53c89d4aee5bfaa17db84501aa4c4eddd59bcee486df5d48d3aa48173d", + "src/inferrsi/data/skill_library/repositories/husky/references/troubleshooting.md": "3948711437d25822d389dd857c72dc7fcc15727c9048d2e3eae3317d97aca6dd", + "src/inferrsi/data/skill_library/repositories/husky/references/verification-protocol.md": "83c5a966ed3f721b412b8d8417af8336c5a8c1e46044a888720936bf2f1deeab", + "src/inferrsi/data/skill_library/repositories/husky/scripts/audit_evidence_card.py": "291e62701ef7a37a5bfab963ea4e20abb94245bc015584223a101f18b1490895", + "src/inferrsi/data/skill_library/repositories/lily/LICENSE": "b9e2b50a48a09fb39dafa71ab1a11a6e87389addda17faa99d55f5f74a0001ac", + "src/inferrsi/data/skill_library/repositories/lily/NOTICE.md": "7a01e666044fb8aca1fd88c636bca43bcc95b8b1fecbe101dce4c7c80fbb5759", + "src/inferrsi/data/skill_library/repositories/lily/SKILL.md": "db815674d5d36964902041d310d3df844d00ab2e6bef0ce4b72ed2696ab2b9aa", + "src/inferrsi/data/skill_library/repositories/lily/metadata.json": "6d9d54e84f957881cc26894eb82ae666f106cfc6c505924f9a6fb6b35eab457f", + "src/inferrsi/data/skill_library/repositories/lily/references/repo-provenance.md": "96182f6cb3666f89adf1aed6e216c0a84ce3a04a0685a4257e286100ba8ee367", + "src/inferrsi/data/skill_library/repositories/lily/references/repo-routing-metadata.json": "f2f3becb51056e03081c7c4140e19572a6318c3d63015884df56eb49d38a00fc", + "src/inferrsi/data/skill_library/repositories/lily/references/source-identity.json": "4595559c11611c855be150919fca3eeb89e7f96c7a8dd133b9f79a5cd8e05e62", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/model-runtime/SKILL.md": "dc800f07fa0e82213483c7b848dea95d609386296006bfec7a34cbb194a83de6", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/model-runtime/references/model-and-memory.md": "d5a862303d646ebbaf745818caed8f96c7323cd442cb6d6e2aae00b0361080a1", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/model-runtime/references/troubleshooting.md": "b32fab02b561c3ab4e683abdfa80cd4a8f929c514147709febaa6b6d1df65909", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/model-runtime/scripts/screen_config.py": "5961bb6f31ceca43473683127b37faee268234040a013ae49bbb4898eacb6f04", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/qualification/SKILL.md": "0231b50b9bebb68d0fc5f2ce19aa7c6e5ffd498d8eef4cc00d39e82908cbbe91", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/qualification/references/troubleshooting.md": "01607534ce1f09fb046255c78f8c1c436aac61874b93b65a9d62f3eb480496c2", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/qualification/references/verification-contract.md": "6934779f67b878bf7cdb5f2cbe461f6716c243c662a29272c6db37f64cb317c7", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/qualification/scripts/audit_benchmark_record.py": "59534f58d46ad438c7bf88886a9b05a8716c922d817f2d9c58ff1b71ae50ee85", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/serving-cache/SKILL.md": "7e062ff9af76125ef08c98b4465e0439c66e9e237f0c39ea0285005cc5496cfb", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/serving-cache/references/api-and-session.md": "7262bb077c24f2f7549d450473ec7944829940d09cca89db9c12f37ac90d5db0", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/serving-cache/references/troubleshooting.md": "6e2f36d2f7ab9eae665d87a676e34c0ebd2f46f3df6fa5f2e589a306749a87b5", + "src/inferrsi/data/skill_library/repositories/lily/sub-skills/serving-cache/scripts/request_template.py": "d6b00c9f0b38c7fb0a0e52e396993a832542ea0aa6e2456026fabc0c39a4d33a", + "src/inferrsi/data/skill_library/repositories/lithos-metal/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/repositories/lithos-metal/SKILL.md": "4c71ba2f26c275829466bba2412738aa27799842e642c75cdfd36d9ed8713f7f", + "src/inferrsi/data/skill_library/repositories/lithos-metal/metadata.json": "ea2d4e73e33a49433a2834d19489653f828e56f69c4975cc32cf508176eb24ae", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/mlx-license.md": "ccfab7ccb2ea306f71531c8ca77bb55507606cd90768b1e32b8b52ab5b48cf01", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/notices.md": "b2c9d2481ef2c9d16b3240c951ae65db417e5d309cd06b7917db9d84fe175886", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/package-api.md": "4f3ecfabafb2a15eddf21ccf964b8c37635bbb7862418a2c404e7049c0fa2898", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/repo-provenance.md": "c0c7c3dd45eab0d55bcf154760156db58e8e365f9cfb2c6bd246dd6a06f3ef3c", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/source-notes.md": "110dbc995c7fc727632017d4c26c71df7ecf12d4df5d89df0cd4afd75c5892f2", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/troubleshooting.md": "1894bbe1a18018f81318b11e2150201ee97f3819f84a7a81f3cf5fb01549c165", + "src/inferrsi/data/skill_library/repositories/lithos-metal/references/upstream-notice.md": "cd80bddd6b6204dbc61c6ac762ed206c73d9264755d9884383d0860de29d185d", + "src/inferrsi/data/skill_library/repositories/lithos-metal/scripts/inspect_contract.py": "f8365f87af2135fcf0a2c8baaefac563c30989282fe080a02d958a37a215515c", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/compile-fusion/SKILL.md": "e0d927763f3774440abddefc45913528946cd29f7f089766fab47b212f9b6693", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/compile-fusion/metadata.json": "b3ff3702531e023b49beded950151ec1256aab4fbf4c02e9212ef0202d3ad72e", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/compile-fusion/references/mechanisms.md": "6173a8b73ec5138cb12791f108f3d5d83278a117d443d48187c842e16bdc6eff", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/compile-fusion/references/troubleshooting.md": "f18ca9324d9a80f2ba327a1ad4ccf84a01049080366fe12410f2800d5f178232", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/generation-state/SKILL.md": "bd8cd27ddb18a04c9e2b31ce9f16e676864133e160267473979de828b1915f63", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/generation-state/metadata.json": "dacba1953fad82b32b98f76bd48009b04109f1ae446cc68be522b02a64306157", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/generation-state/references/state-machine.md": "d812ad31d4e343539119cf82f8ce241fe14c691dd73398048893cc5de3ab8e22", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/generation-state/references/troubleshooting.md": "7ca629e66ef3b0a67077a7d2536befef683315ad01939728a9e6e7523cc8414c", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/serving-cache/SKILL.md": "01c38c2678d6e8292a3908bf5619b332432c5f1289fd5d5a67eb44a2141adb90", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/serving-cache/metadata.json": "2fca222efac7752e29f236f40a0bc38275045826b55955687e39c64ffb69f58f", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/serving-cache/references/troubleshooting.md": "52c6e27a76547475b5c66121f873f4d2c1662c811a69da64ce85614f1bd44749", + "src/inferrsi/data/skill_library/repositories/lithos-metal/sub-skills/serving-cache/references/workflows.md": "298d27d3c748fe22774b872ed0465b707e53812379a7a538c95f25d1727fc796", + "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/SKILL.md": "95e81b6bc096a954fa73139ea5bc001d1efa6c99747161da62f012b494e5c013", + "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/metadata.json": "1d30b31dc8352fc6a8d7b8abbef8118c840e47e5b90795dae0c758bf9717278c", + "src/inferrsi/data/skill_library/repositories/llama-cpp/audit-llama-runtime-state/references/source-notes.md": "3d12a401e9decb55a2f4f3b2bfa0d360ef5110f63a0469dfe285654e78054190", + "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/SKILL.md": "e4b78f368d70c8abb74a935acb910414015f1aeec1b50f7a295e836822ec4b27", + "src/inferrsi/data/skill_library/repositories/lmcache/kv-tier-management/metadata.json": "c221189b8d85fac3bd41ef1c7cd560f1aebc8cfa0381c182d1d571a7169be829", + "src/inferrsi/data/skill_library/repositories/local-dream/SKILL.md": "e51fd3c71d693070abeba57ca67dcca82c7347cb94eb98f43e7c455bcacbaa46", + "src/inferrsi/data/skill_library/repositories/local-dream/metadata.json": "ab33602252d0f5bb0797c3e1f12466b55a82aeaff10cd24e42a99a5224c8e593", + "src/inferrsi/data/skill_library/repositories/local-dream/references/android-lifecycle.md": "f85e5fce39f5de5f88548fd8c21978240c810f899e03ff05ed72eb3508ee5e36", + "src/inferrsi/data/skill_library/repositories/local-dream/references/native-contracts.md": "40206892f5e421f9d8edf303870f5c4aefa3e87123537178d45f438f3d04b7ce", + "src/inferrsi/data/skill_library/repositories/local-dream/references/packaging-and-abi.md": "9a19bca59e19c3bfbfc1a15f5fc756c13d954ace367812aeb000551f4484ad06", + "src/inferrsi/data/skill_library/repositories/local-dream/references/repo-provenance.md": "0b848082c0ce893d8ae26d779c5b663d914d1a208cb2293c4a84668cdde86293", + "src/inferrsi/data/skill_library/repositories/local-dream/references/source-identity.json": "b85f028bb96a37852e9d9780b5d5052fdd9c8f3a704ad391b6381adb58291347", + "src/inferrsi/data/skill_library/repositories/local-dream/references/upstream-license.md": "41003d4a74749c0220e33dd415042164b5a1093ed401f36277234f772d22d3d0", + "src/inferrsi/data/skill_library/repositories/local-dream/scripts/screen_declarations.py": "ccd7d0d55be9a47484ec3a74498d8edef8aba55287f92287775c8d82050c8932", + "src/inferrsi/data/skill_library/repositories/mistral-rs/LICENSE": "b40bb09c7b3b6f697a8920cf9d2cb3ff8cfcf77a0fcb6d264a92842b43c4dcae", + "src/inferrsi/data/skill_library/repositories/mistral-rs/SKILL.md": "8917b30fb86437dea7508275ccce1992194d6e0c4858d537a410bf1f900953d0", + "src/inferrsi/data/skill_library/repositories/mistral-rs/metadata.json": "5e88e2ca64b848b8177336351e03d1fbcb5c495fd54217daa83710ef6faccff0", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/apache-component-license.md": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/paged-component-notice.md": "a1718d8c5a25f15e61bea98a158880e96faa5b7174e82c4e1470fbacb23a3559", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/repo-provenance.md": "fab87dacbe0a7298e42137cab7409f443ece0dfefba60961040031e36d96adc8", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/repo-routing-metadata.json": "0debda77a8f6b1b1c06176e3c2447767c1953b9fe7e4c84ef01b3ac90735df70", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/scope-and-notices.md": "938203995f5e58cd6cefa278225b2d5a3ea7dc7f07cae86201f9fc71553ac020", + "src/inferrsi/data/skill_library/repositories/mistral-rs/references/source-identity.json": "6a8c0b0834228b3a298e7e548c9d8cb774c00f661135f51c12ee613efe1decf3", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/SKILL.md": "73620deb4efad4681da10a07bb982e180423bd69fec3572d16ab556c530b89ec", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/references/helper-contract.md": "802d8af65343eeb29b0950d3c305d0f24cb7a4298628ade5f7cdcf4358c9e228", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/references/layout-and-features.md": "6804a771ed0dd2d8d60613d52ef87d4297af2a6445b4fe61d08adeab4be331aa", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/references/troubleshooting.md": "a6de920f174d6918a72093921d344c700108b0c01e1f0d09f2b9051800b1fa7d", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/references/workspace-and-evidence.md": "7ac5028be6fd48f9c6cf067bf832efc8edd681195f2fece67ea119211bdac106", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/decode-contract/scripts/audit_decode_metadata.py": "de28f586f27278fee25f6c96f1e00ee63528cdf746f57c83bdf367c31c0f6e6b", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/SKILL.md": "21015790789933e106f6ea929fd1778d862315520b55425213c679e9b7a44e95", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/references/helper-contract.md": "607449105bf3bcbcfe785dc9448a7a587c7daeb528ca63e0cd959b074a9bc409", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/references/prefill-and-metadata.md": "e8a738a668a23068b8ff1907d521a3a26f7090e7a791f3a715efdb6001338dc0", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/references/troubleshooting.md": "f584a02528e2249f5d12f223d8933e9d40212e5aba8dbe545261164b00ed6f8b", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/prefill-contract/scripts/screen_prefill_metadata.py": "769366c19d621ec1b42c722ed224fa154a63c3410a0cffc9072a13abd9c71bbd", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/SKILL.md": "745d88dbf6c764f64f8b31763e57bef720788830ddf17b518456323000325fbb", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/references/prewarm-and-evidence.md": "6b72ae4b4599018ce2da5c54d0c28fadcafb7ce75b81ce4c18c06a440aab6056", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/references/troubleshooting.md": "d933bea6e9e493faa8cd0f4103ea33197ace7f7f901359f8cf0d6bab2edcdb4d", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/references/warmup-inventory.md": "1e7af7d63c6853ede4232b5a12cf68aba6597c4a305015366f1c2c917e3caf01", + "src/inferrsi/data/skill_library/repositories/mistral-rs/sub-skills/qualification/scripts/warmup_shape_inventory.py": "6aa920ca1cb534dec41f5dd2d833c3e760b234ab2c6439b6c4250b8a30b74055", + "src/inferrsi/data/skill_library/repositories/mlc-llm/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/repositories/mlc-llm/SKILL.md": "e0fd1accb4216191c56e7e4a8074633f36d9063727c2546e0d2eed14220d8cf4", + "src/inferrsi/data/skill_library/repositories/mlc-llm/metadata.json": "c9927e284e44aa09733424c024409ec8efc9593520418bb7d2e53ef301363681", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/mlc-notice.md": "3a7dfb37bb96ce6f5a1cc0d6d4c8da3b8a35e6ec4692647ba0a37f6155c7e694", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/repo-provenance.md": "423f21e1de6d7ab2e11c0bdf540357efde025ae4535bdf76ed9770b1a59ba35a", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/repo-routing-metadata.json": "5309c27e7b60c4fe7aaef28bf6911e7e3c936b58b1a907aa8486bf0e40b4c0d0", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/scope-and-notices.md": "8c573c234f44a0808a731774921fb5228fc0844cd238496a2c4d61d0120bdc79", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/source-identity.json": "75e91b0f7fdac12466e62eedfb49baa6f1b3016111d7479aa2b04e5c799454b2", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/tvm-license.md": "e62264b483aca7a9f8c1ac69388b476971b43975ced90ac11dbd9e9c530edb46", + "src/inferrsi/data/skill_library/repositories/mlc-llm/references/tvm-notice.md": "967cd2fb97712c43271e2b1c4f636c4d8823898869f0f76f65bf8a3232584f0f", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/SKILL.md": "933fdd05ca853f47e3ace024cf1e08df7511834c31676851c28803b6dcb1c681", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/references/contract-mechanisms.md": "82187f4a1478352c68a5690101e9798973ef63f9b6d717e06188b03a1832bb8c", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/references/helper-contract.md": "48492a36435ff1340a4db4e0959f052cd72683ced26bb8ab35baa84bd79ccd7b", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/references/troubleshooting.md": "41b002a31b031a78ca372504de4c9cccc9353d6d80d989b8d62e56260c291e2e", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/artifact-contract/scripts/audit_artifact_declarations.py": "874a16cb16d75a14cb51b4fbe442f4fac46feb21ac242063f5f845a0fcbabaa4", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/SKILL.md": "25dbd48f4fe84b7d759005d04a2aa9db33da987cafac7975a148e62cc9efae34", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/helper-contract.md": "5428efd6dca0654a58e9e226805e15c9a9f1b3f1ef60db234139985d107acbfa", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/modules-and-streams.md": "56faaf149f95ee821c242ccca8115a6f3b166d34392499798ae8532fcd10372d", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/native-evidence.md": "133322e879fd9010a7513afd278d937838aa2ec97d593ed11d324cc59793374e", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/targets-and-export.md": "547eb8e56af4f7b58415eb58c338ea1a33ff8edf8b58fa523700947dc7e04821", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/references/troubleshooting.md": "e392b8329ed183bba5fd79fbd260c3fbfb389016ee71880a957741e71a0d59c0", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/metal-build-runtime/scripts/audit_apple_target_declarations.py": "d5bf16f7def4cfe28b143a700ca45a9636ddaeac12ead04cbaecdcc8b4b71aa5", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/SKILL.md": "6b4914cb60831c07d4c748162251c49fc16b229e0d62b4c5a5663719c45b16de", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/references/cache-lowering-and-evidence.md": "a118f5bc584ea885e8c573978eaf1e32d448910860b8135196c5c1acd42aee05", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/references/helper-contract.md": "da33306472cbda5a33605b80fe52d6cb1fa886b3b7eea13adbdf0ce8a6024c66", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/references/limits-and-workspace.md": "77105db95de2b8ea27ed6953b5441c8af061fb745ddb13d3904f8773fc21e879", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/references/troubleshooting.md": "b89255fc04507185bc8bcf74de779730b009b12d13bb38ad0193205dfed8e921", + "src/inferrsi/data/skill_library/repositories/mlc-llm/sub-skills/runtime-kv-limits/scripts/audit_kv_limit_declarations.py": "d82a6dda5f5931eb369003a5e1f419bfbdd82304766a18be8ea1e1849e26b77d", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/SKILL.md": "bf1177cb9985c12a9bf11edb64771018e1e314c4713920de8fd4a989ff70115f", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/metadata.json": "d86cecb615e93ba0f6be09d1ead102cd5a17068c07e1be2549ac5cd8f7519279", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/cache-stop.md": "bba6b5c73261065309cbcf11b1e591eef48916f0c79ceaaf74d98a9fa96b0992", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/generation-batch.md": "f16d3c57d45e63bcaffd3871aa41c76972ee11f7dbf1e12b45d9b170303cad9b", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/source-map.md": "91d26181f246307f2113f94b12ac9587d689b654dcd216e9f14b11cf5ebb650c", + "src/inferrsi/data/skill_library/repositories/mlx-lm/generation-cache-lifecycle/references/validation.md": "73b5dd347017baba519ecc85415aead9993c1751d116b9ba4f07689f58c5f783", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/LICENSE": "7b1d86acac816ce8192209d7e01141c129b4d48d474c15cc5fb4af62f497c323", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/SKILL.md": "501f73e7b327b04b6b80501b42b20dab1ac15c875817e6f4c4b66fe3e69b4fc4", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/metadata.json": "014131f345a28d3367313ed5a4cce858fe4c93d4d47a5f03eb89f609592e032f", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/references/repo-provenance.md": "f1b15edf89c597db0c3fc0bff7db6b2451924ce9273c39623bc0cbdccb568d31", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/references/repo-routing-metadata.json": "8fc804347771cd768a7ed027f2041d96c9832fe5f8c5869e6593a3021b1bef8b", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/references/scope-and-notices.md": "18e56805a8f275d35af3d8e42f3aa1525871b43400921a5d2dca4897c2047214", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/references/source-identity.json": "518da7bb64f22ff8ee6beb0e0f0579e7cd0da2fd70fa69fd8288294b0c08ba24", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/SKILL.md": "04b97959b4e207380a06d932e9a557954a9fb9aa880c243439da6d4a981854ed", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/references/helper-contract.md": "1358531257e00c0f9fb2eb9fb7a40f9d749f5435c90168c8923b6646353dce6f", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/references/recurrence-and-state.md": "ff0ebbeacb8219d93a890418d18bdf2856e6326d67f974a7050de3a0e5a3c5a3", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/references/troubleshooting.md": "3bbecbf91b241e7b37b3e8c753bf3fbcbdbba0b0f4d0ddf2f65eb653007f6cff", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/references/verification-boundaries.md": "21b6fceb692f29b410dc967f82e31d0d082904cea827443ffc05f6f940191dfe", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/gated-delta-contract/scripts/audit_gated_delta_declarations.py": "3a256c37c867866e49dfdc2893cbb3f32d061e389fe4395555eff899bc0415ff", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/SKILL.md": "7effaa20ddaf32e3b4008c95888ccd09bc2bbc3bb5ff17c710f989b36be49535", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/references/evidence-schema.md": "d40e44910d8fc43ee3c731735902b1c1ba506c2ae972bb53dbacfaf37e38509e", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/references/native-evidence.md": "3e9a630a30932989faea5f52b347318cb94c73c7d4d6c81c0465ccaee2818671", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/references/ownership-and-termination.md": "56fdec7ebc89b5c217c7e5c203886bef1d3c2eb5794852897bdb5941ab1db2a0", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/references/troubleshooting.md": "de28b8ca9cf95be2a4e5762f938acfadb6d4b53459dbf86df4b2b2ec64ec347f", + "src/inferrsi/data/skill_library/repositories/mlx-swift-lm/sub-skills/generation-lifecycle/scripts/audit_generation_evidence.py": "2aba6caf42ca1b3fc55299fd96419af8681c37750c21a7ba75205a43baa9c7b1", + "src/inferrsi/data/skill_library/repositories/project-maya/LICENSE": "574cffb7c57769a9271043ac0c2b07c5e7e7621cfb81081b901d8981e791d121", + "src/inferrsi/data/skill_library/repositories/project-maya/SKILL.md": "c619a3472be02147a847a24ca97c56afd8b56b6ffd22ce66fcae453155e34bab", + "src/inferrsi/data/skill_library/repositories/project-maya/metadata.json": "c9f7879ca512c07e9bdf6d1d1a52de055a0fa210b8838c205e27a34436324834", + "src/inferrsi/data/skill_library/repositories/project-maya/references/license-boundaries.md": "090839df90d6e96f55794c2ad821d7154dc6a6aa4211cdbb08a1e5f6f2e52a92", + "src/inferrsi/data/skill_library/repositories/project-maya/references/repo-provenance.md": "7bb6b6c14d1a9b77e906804505c723569f952e5d47559e636d2e6621994216bf", + "src/inferrsi/data/skill_library/repositories/project-maya/references/repo-routing-metadata.json": "0acbb111e92316e74d537200cd033f3d7d0f69972bf0953c8cea9dfb2ecb4a34", + "src/inferrsi/data/skill_library/repositories/project-maya/references/source-identity.json": "fc2287d5bde587cf040bbf515161d25e84a7bb42989cf01343582af67e4b8eeb", + "src/inferrsi/data/skill_library/repositories/project-maya/references/troubleshooting.md": "735d4bf3a3b243c29e6371af37dc06c1fa43b39a85811ba9714b911bf520ff67", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/SKILL.md": "1b68800cc2dd9b7b03cee3343fb8619135dc3563550cc20386606c295358df6f", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/references/model-and-memory.md": "ea7b905fa72cbfda5513f4c6ae77c552044af71e9313e396bcd3628884183bfb", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/references/notices.md": "6f6d9a96e1e67da6fe25f54482da0b69bd256fb688cc0956668f7319308f310e", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/references/troubleshooting.md": "97c00435d57a5b9067fe8e5885219d5818208a25f225a14c240d7160dcd697f7", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/model-runtime/scripts/audit_runtime_plan.py": "b096ec10c4f0e4b057997fe7b946007ab4c06dd263ab09d803e29e498601da00", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/qualification/SKILL.md": "5e264000198a9bbdded9c35db2df682e8eee4408ab741f18bc002e9be103b401", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/qualification/references/troubleshooting.md": "0c42256af4869e5cd1eb88fdaacfe65adf51dd180bdc03a2b9fccf220a19ccd8", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/qualification/references/verification-contract.md": "eaeb9319df27b66840cdfcba5c4788a287775d1493308517d9124fd0608eefec", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/qualification/scripts/summarize_decode_records.py": "cc0ac3c403aba79a131f631290d68e1640283e9f0d564aeef431b577824e3036", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/SKILL.md": "830d8bccf6b36e1ad6cef899d5147afb74c0514516c68d02060bdfb550b56920", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/references/api-and-history.md": "a51fa298017f266decd801441eb411fd68f7906a90a4b0d18083d1af31fc7d52", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/references/native-state.md": "e5a7ee5e777998db0fb5422cf5b1ab18dbd67b4cbbfcfd6628ed6e40d8709c61", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/references/troubleshooting.md": "97dea7b2d26fe165d2f64a1a63c91d099dddb2d3b95c16b667f5173b3ffa9d2d", + "src/inferrsi/data/skill_library/repositories/project-maya/sub-skills/serving-state/scripts/request_template.py": "d91b03fbbd7003012a92bd7259d39604b9a91ee5707da2d3366e7c0fb47edd6a", + "src/inferrsi/data/skill_library/repositories/sglang-omni/LICENSE": "a54c8553df5d4313d293d1aef3bf2f13952bda8b7c7a1fdd729b2bbe27ee6015", + "src/inferrsi/data/skill_library/repositories/sglang-omni/SKILL.md": "d75f5b2881b240ec2d9d4f5cc8e3f7d53e4378cb01732a2fb9699da2901ba38c", + "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-provenance.md": "5e8dbe168a8885e209d4b4532944ffa7fead658b23fd6588054d57b22476f63b", + "src/inferrsi/data/skill_library/repositories/sglang-omni/references/repo-routing-metadata.json": "6820707cb9bd877d1635928cf5a0efe38e4fe37dc04036dc70d53e479ed9bf7c", + "src/inferrsi/data/skill_library/repositories/sglang-omni/references/snapshot.md": "0fb5d776349de845dc314f16ce73a06c9caf5db84dc1c293327d3b3ca1337c80", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/SKILL.md": "766b2d66d2056dd779abd9e492cb3af6e25899298f44964f565509f130a7cb24", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-architecture/references/runtime.md": "36bba9874fe054acc5217f9c265f5ae309c83943e0d46228370ae1bcdc69421e", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/SKILL.md": "9af38bca840d0004fc2ed049141476fc40fd40dcfa843f38a404a7a314c19425", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-extension/references/integration.md": "6ce9383ca17bacff60a180f8fb8c43ea47eaa37117fb8da6051ad86135d228ef", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/SKILL.md": "990b6bdebfe22612de49dbebcb16810e8abef94d70691aa883b7937c85210eb6", + "src/inferrsi/data/skill_library/repositories/sglang-omni/sub-skills/sglang-omni-serving/references/recipes-and-tuning.md": "94afb762bbe185f58606154b3f0a5af56e7089cb84ea94e26cfcb9f93c95439f", + "src/inferrsi/data/skill_library/repositories/sglang/LICENSE": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747", + "src/inferrsi/data/skill_library/repositories/sglang/SKILL.md": "f2b68d4cac90b5bc4d0e38cd17b9f5bec862e934668a459ee08f69a21b5fb72d", + "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/LICENSE": "1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747", + "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/SKILL.md": "b0a4ba4751087126143ef86983798deb0ed00e24b6ff06bf6ef76a9df6e274f2", + "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/metadata.json": "6ed57dfb4f25ecb34b81f2188b055e05d172ba2581e6a3b605b814b2f733a4d6", + "src/inferrsi/data/skill_library/repositories/sglang/audit-sglang-diffusion-lifetimes/references/source-notes.md": "d0fcfd1177ae2f351041c5b2abba7f5121cec133006d9e4dca8f74ddb567332d", + "src/inferrsi/data/skill_library/repositories/sglang/references/architecture-and-memory.md": "798e747022b7ec135cc4861c2c9ac44fb14b503bb77a0068ee7fbea494b764a7", + "src/inferrsi/data/skill_library/repositories/sglang/references/native-candidates.md": "abafcc9a34df196a6365145554887b8d37afb36c6e236c92e50a4a06c8a4ec93", + "src/inferrsi/data/skill_library/repositories/sglang/references/parallelism-and-extension.md": "e3be7f4b51264ac68682a64d264e4d96857d5bb9f9d99cae5f327c56984a9e6b", + "src/inferrsi/data/skill_library/repositories/sglang/references/repo-provenance.md": "534d95ae8d0b29bcbc8f6707b3d38fb964e31f5b66ba13517a9a6b8f309bbb38", + "src/inferrsi/data/skill_library/repositories/sglang/references/repo-routing-metadata.json": "b2b879c1cc42ddc2a5b95ae24115eb891562fa9a6e96d151f4ddf6f7719fe73c", + "src/inferrsi/data/skill_library/repositories/sglang/references/source-map.md": "c2adad64ceb0c305cdc4859c03012b3e0a4ab26cd0cb289bbd76e192b0cc3794", + "src/inferrsi/data/skill_library/repositories/sglang/references/tuning-and-validation.md": "49b3343b3292cc098a787de9d24d968b444d0ef4f73fedb5d8183390e8f39c9c", + "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-parallel-extension/SKILL.md": "18bd54ebdd629e1ba6dfa589f391c7cc3c5355785c2ae82f741b64ee8022464f", + "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-runtime-memory/SKILL.md": "33b1a65aba00fca819d2b6cc42202e7708e9155e4d8f4373a2387f4ee2f2fa3e", + "src/inferrsi/data/skill_library/repositories/sglang/sub-skills/sglang-tuning-validation/SKILL.md": "744b2ef9122101d3f50cbbb40fe020e2b3d3888adbe2d86041041e7c7ce424fc", + "src/inferrsi/data/skill_library/repositories/splash/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/repositories/splash/SKILL.md": "619b87c282dcec4befa319085764fb53fa506a05c6a7112500bcca662f059ece", + "src/inferrsi/data/skill_library/repositories/splash/THIRD_PARTY_NOTICES": "2fedf1277bf732f8435c3e3be2798f52f25f7e2bd4155865cb366da55b88e50b", + "src/inferrsi/data/skill_library/repositories/splash/metadata.json": "d4c679f474c31549156aad417e23c2802b9f4bc904472e5cb16916f2996b50ba", + "src/inferrsi/data/skill_library/repositories/splash/references/repo-provenance.md": "20e3462f1b6a8752eb0fa6f36e76867f142a49c1953f498f895e29c4fd4f0204", + "src/inferrsi/data/skill_library/repositories/splash/references/repo-routing-metadata.json": "287ae591f42949b8d33312659ec8303c47d5c73093c64c125edb1cf6ea8de8b9", + "src/inferrsi/data/skill_library/repositories/splash/references/source-identity.json": "2f778882ec440538a46f48332650412b446722246c8dc5690083f19da7dce61d", + "src/inferrsi/data/skill_library/repositories/splash/scripts/request_template.py": "086d773d42edf67c43cedae16974061b09ac66458ee47fe961fc01733f3b2a0e", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/qualification/SKILL.md": "4fddc39811c12ec4b3279ad68a8864b7f1ef80e838ea03788b54cb025a03ae16", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/qualification/references/troubleshooting.md": "dbd2179a48eb3fc0ecfd54c5df8a690d0cb2e1814acc19ac3b0aabf184092c92", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/qualification/references/verification-contract.md": "9efffa0ecfd3b6771dfce631137eccc6d939c3239f60b1fadcd0c6c1ed7700be", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/qualification/scripts/summarize_rounds.py": "107a844954e3909500c60eb1ca365ac9b5e7dc3d91122dbae46f402a5cef4e3b", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/runtime-planning/SKILL.md": "f0fe3d3f4053937767ae586559cc7fba89dacb6c896186a029fff220414e8d22", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/runtime-planning/references/memory-and-formats.md": "04edf10f4a965c512f6a262b1f8295f3c033d615849e49d21d3d94c78b549f2b", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/runtime-planning/references/troubleshooting.md": "5a1896f9fb01f6c8c9ef78d94f3c965d6d64e6b54223c6d7901eaed0548c5575", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/runtime-planning/scripts/parse_budget.py": "b3a3b0847280333c2bfbe5d03350dc866307d5290160ec64c7e3d462a34f6301", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/serving-cache/SKILL.md": "c16f911498e116ff9b31b5a6f0ddd3ae6517b51c0e9d079c337f36a5bcac1df8", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/serving-cache/references/state-and-api.md": "592603177473ac8e024b7dc19a3d3faedd7c1e0312137eab8bf9a62d2f567be6", + "src/inferrsi/data/skill_library/repositories/splash/sub-skills/serving-cache/references/troubleshooting.md": "4fda461042a3de43c109d7a3cec67542c7d61fe68412bf91606220f2e13e576e", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/LICENSE": "b53fa08f515cb5a6fff7b9fd8fcd0961a4b80df29d74372d25e1f9171aa042ee", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/SKILL.md": "4c04d39812fb566e3ecb5807c170d5c32848eb2b93a1f0fedcf66ba9a9c92860", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/metadata.json": "13ab6d874c4e3135be80ef7c3cb6bfd28fb9121012e0536831ec3705ccb8c5eb", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/loader-storage-boundaries.md": "c1ec047aa37fca6b374f9bfd750e8f30d45af1d48fe728db23873c0b84ad94e8", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/phase-residency.md": "7b537cb972347dd68e303960f142892346d8b19b884e4d08016ea36a23bdaa7a", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-identity.json": "af4830370b6eddc2d176a9ae14da4ab56a90b0fe048a01a89ee051df4365d8e9", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/source-notes.md": "c493af8f6fd11ab7c9a37c9eec78cac31d57c04403a790035efbb4fff99f4669", + "src/inferrsi/data/skill_library/repositories/stable-diffusion-cpp/audit-sd-cpp-residency/references/validation-boundaries.md": "3aaf79b36192bf8ff3e8f95ac90f2e769aa065dfac1fe70368824f3591849b8d", + "src/inferrsi/data/skill_library/repositories/strata/LICENSE": "949b9b0aada7476006469e58ee4ffb0e0f68c2d9137e90a356757b61a09d37f0", + "src/inferrsi/data/skill_library/repositories/strata/SKILL.md": "0e5daeb6409c916be45ecd7ce2531ea5f1c60a4a53e2b0ed25dc1e2cdc5c17ed", + "src/inferrsi/data/skill_library/repositories/strata/references/design-space-analysis.md": "11ecafe71530ddd4839268ea1b24b3361a3895012a705859ca06dae29fa72630", + "src/inferrsi/data/skill_library/repositories/strata/references/repo-provenance.md": "05f17d0ae996ad6c12314bdd29f3427574f6aa611ea60effd18598a0f386638b", + "src/inferrsi/data/skill_library/repositories/strata/references/repo-routing-metadata.json": "a10ba49b7e6565109b566c105e17551d042075a1ff649c6f371134e3b7402c74", + "src/inferrsi/data/skill_library/repositories/strata/references/trial-format.md": "92d8cf38549268ae4eef382890662d03e18a7dff858ad45ee5eb6f47eb39c8c0", + "src/inferrsi/data/skill_library/repositories/strata/references/troubleshooting.md": "7301cfa5a5f6cfbf9576c3845a96c6b4de8f8a6509f2330970a8e2e4669b3d5a", + "src/inferrsi/data/skill_library/repositories/strata/scripts/summarize_trials.py": "6a243a08da15659ced6519f49f52a49a675bd9e952740c3ab03dd21964eefbe2", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/SKILL.md": "857bea3c64db45dc035728a41523142ac65778b41f63092148b8ab936f971d76", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/design-space/references/design-space.md": "0a428dcd12d5290b554e9da6795a110f94d98f128771602e1ea5f608048b135e", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/SKILL.md": "cdf51c537e7237296d52ef6cda9f2e4c9d9ce3fc87265a369cd9eb7497c53512", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/tune-execution/references/tuning.md": "7b69aa7872d60dd3fb91e41b77a632572070dcf80dd5b5f6dc1722d8540177b1", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/SKILL.md": "90bfd975e25f18895ce0e85f6f35d24149ce026305a4d479aea4924510053460", + "src/inferrsi/data/skill_library/repositories/strata/sub-skills/validate-engine/references/validation.md": "dae846f4c47c68731b96ec1d50cb008c03871c72078f1c5b0f4a9423be9ba397", + "src/inferrsi/data/skill_library/repositories/vllm-metal/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/repositories/vllm-metal/SKILL.md": "5f83b1b989e8dc239a6f34a6b3bd63872533ecc1404483a69dabfb1865648439", + "src/inferrsi/data/skill_library/repositories/vllm-metal/metadata.json": "63d51d1ace6b0ba503c28751a8ffbf9224038395f4146a94c775c628d8075470", + "src/inferrsi/data/skill_library/repositories/vllm-metal/references/repo-provenance.md": "970bda4d1081704aa82c94d6544a98bbf479430c4fc2c82a1a82f206043aa72a", + "src/inferrsi/data/skill_library/repositories/vllm-metal/references/repo-routing-metadata.json": "230a2b383bcf2855af4de50374c41f11b0b0841268ae45e424c1c97f679b477a", + "src/inferrsi/data/skill_library/repositories/vllm-metal/references/scope-and-notices.md": "aaef98eaea4cff56f39852e1d131163ebc01334e79034b1890f67a2660727d3e", + "src/inferrsi/data/skill_library/repositories/vllm-metal/references/source-identity.json": "9da529574e81f9f91d86c01153eb09a8b1e728dc5327c40364fff74085598c1f", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/SKILL.md": "dbfb06896615dc994f4a0a1e3789904acdf512ec316cbab6347165b764bc592f", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/references/cache-and-offload.md": "9829c14cf20f22575c44c1a0fcf18f63578443710e1a592c9fea709a43a3b888", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/references/helper-contract.md": "56a7494b58c47e07d1c85471ddc6075b15e34dc7a852b937002416dbd55b65ce", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/references/troubleshooting.md": "a4600c6d17a980b6442fc4c36f1c9ae148a2710c644126a9a898675b01f95366", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/cache-lifecycle/scripts/screen_pool_budget.py": "dd615f7d3239f48510ce83416fb14fba8d5da2b9017627bdd312567be1bd7956", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/SKILL.md": "e2f5c5207b5c7665d3388d94fdace35e25f073a7833cbddaf14514a05f5ca5b9", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/references/configuration.md": "69949b93f4f319145e37d1b28487e37bd4ef21622f3169583d9cff46234a9c76", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/references/platform-and-loading.md": "7f11dc9e4bb9ae7bac25fcd6a2909cbea0eb2ae329e28ec9a7a2f4085c56e5cc", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/references/troubleshooting.md": "837cef82321fa050af3f01804fe337b0a0462255261fea5c21158ecfa3d0c325", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/platform-config/scripts/check_environment_plan.py": "2603200fef87e3afad7820b84994f7e43e6041b45c17d0ed613d0d6229535bf2", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/SKILL.md": "37d9ae55ab4d0afd4ff8053c1955445e5a090188841ecc7941a5e22c5dd5eb39", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/references/dispatch-and-measurement.md": "360719d4253ad7e5a36823171d432aa8ca0f74d1ae565c7803af1990d99f2af9", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/references/record-schema.md": "f3cefb8b131613535dcacf30d130ffc2dea991acfb495adb4e1c7050046903d7", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/references/speculation-and-pipeline.md": "64faced7f31d489e9b0c4b9523a803ff1a76dca0f9315bdc6d0b83935bd338a4", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/references/troubleshooting.md": "a2e85696120ffc2af4249ac9c1e1b96320e02287e5099577c0474c025ab4f5f6", + "src/inferrsi/data/skill_library/repositories/vllm-metal/sub-skills/qualification/scripts/summarize_serving_records.py": "78e660a13a6a6941d006cb11fc2f401a34f489e54273860ae6a208dca6a6a3c2", + "src/inferrsi/data/skill_library/repositories/vllm-omni/LICENSE": "c71d239df91726fc519c6eb72d318ec65820627232b2f796219e87dcf35d0ab4", + "src/inferrsi/data/skill_library/repositories/vllm-omni/SKILL.md": "df3e63dc28de2dac2aa92824fd88719159138650d96487332e0effca3f0e2a80", + "src/inferrsi/data/skill_library/repositories/vllm-omni/references/evidence.md": "ae0d51ba56b34ed81fba34de8fd3685bebf2f160b188fa53719936002881d220", + "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-provenance.md": "43bbfe431434ca4ad6157ba35cbb9ce2d89e4065f427abfdd5f372f51ba0c9c9", + "src/inferrsi/data/skill_library/repositories/vllm-omni/references/repo-routing-metadata.json": "7784b4d98d096718c4f4cc42ce86d5ec113c469844e12de4690b0fccca59d329", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/SKILL.md": "fb3ff50a7b9122a2545a97ef5179011237c4c4ca214dde7236ad7bf79a7c8623", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-architecture/references/architecture.md": "db845eb4207ddeedb8ae381f1b91704a42cb53ba0ee6c7ee581510ba1a412190", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/SKILL.md": "8ab3209b838635395b22deca6aa879172016531d45c2374eb8e498e7f71bb40f", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-serving/references/serving.md": "9bd5eff300d9755137a7600c8f7a7753bb22d09c8853e08062d7e535662ba8a3", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/SKILL.md": "93c83bc4b799b32f50ff54b870cbbde050a790af16d9b7b062039c4987c80148", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/abort-acknowledgment.md": "e95ae5b2cc410917ddf5e9066890d0bef9264104529bdca5ac12bdf56fe80ae3", + "src/inferrsi/data/skill_library/repositories/vllm-omni/sub-skills/omni-validation/references/validation.md": "1542061f9c76fff2acb6529a58238301484b0e23ec0c697d2fdf129e200ea973", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/SKILL.md": "a468d75f25e77b96e6fb5376692f7a89c0464cc00d4df035f1dc6cc923f713cc", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/metadata.json": "db0b4593d7a6fbb898fdf11a1710b529abbade847ea57319843669b375b68fc8", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/references/NOTICE.md": "cca794656897ecdc37f983f7f23c4cd12306dbe3c9cd2f54fece6340753ae4cf", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/references/source-identity.json": "5e52db0a0bda4673b9aac46a3fc035b824780def72921fee6796edb2ed9fca65", + "src/inferrsi/data/skill_library/repositories/vllm-rlt/audit-recurrent-kv-finalization/references/source-notes.md": "0956f59cd9a1fac27da53228dd60c51a07b87639a43c09bfab296c4aeeceaab9", + "src/inferrsi/data/skill_library/repositories/vllm/LICENSE": "50e6751797c50dedd75ef1b8a0d9e42f5f8472e9fbce91f34718e9f97b0c780a", + "src/inferrsi/data/skill_library/repositories/vllm/SKILL.md": "6f36cee8293b5484fd387a798cd450ad10cc2dc64aaded529e3a71ef4ecbe45f", + "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/SKILL.md": "4c3f27459df239f9f4fc971c559d4c9ddfbfebd70507a8a93be8eae59a30b69d", + "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/metadata.json": "b43fc3a7ccae59c664fcdf9ff77962edcc07c2a3df8b59324db551642bd23e23", + "src/inferrsi/data/skill_library/repositories/vllm/audit-vllm-kv-lifetimes/references/source-notes.md": "3665b034594fc8cc767c408002f4e16f8803369242405027c7f37ce6c9ecebe9", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/SKILL.md": "d99fc28d48127d3bc627f2a0cab6f3f16158a77da49428838a08865adaea9905", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/metadata.json": "78459416734f61a8faa819d9a1a3ad7b9a985e8bf8e149813c604d15be6bb92f", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/INTEGRATION_NOTES.md": "82b24be82451ee121892f61575d0cf4ecb9920813a25ccb26cce2b45dda6d8b3", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/migration-is-not-free-memory.md": "2ac168f2725cba7fc92dfcc67c01bf2258dff6a2dbe1ddb063bc60c175d82d0a", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/parallelism-change-with-state-cost.md": "5accc81086e4500d66f894ef6af9951e1cbb76941f6c631aab5c5561a786d8d8", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/cases/preemption-under-kv-pressure.md": "462200a6d0b193ef01f32374a463e0f2838ca8edf33d9ab898def65ffd0b490a", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/knowledge/scheduling-state-boundaries.md": "d4ef29869959b835a5dcb06ddb2fd8e6cff49bb176994dc1d46eb9a852f6228d", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/fastserve.md": "9b4e275843000ef453d508d9c3c9c3f8ea1de5545fbd01d7abe74baa8a4931a7", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/llumnix.md": "f252ce33ac2db017701ca1d49813c5ec1996c4c92900cf35bb6bda7fc59c0f18", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/loongserve.md": "7bc6d1e256d461b04acc8ed856cf99108faceda9cde729ab95bcca596fe019ca", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/notes/orca.md": "f5a3b5870d64226408a8fcda5d69483769d823f8a5efd7f11698c8da483ab195", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/fastserve-readset.md": "9866bd0c78b2b638998b3b632f6bae26aef032462c81abee336d309af5b07831", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/llumnix-readset.md": "3a2819d65ad24087c08530492ef39d8ea046ed1f780d53512f1d23e2b3578487", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/loongserve-readset.md": "f1a383e7bdbbd5a5a472d7f6e1292c8a17d4910ff736ca5a6aa422e1bee74f8c", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/papers-20261015/references/orca-readset.md": "a497ef23b7e60c0fbc15d95b7e07cd20547cfe56f37c24dfe6ed70d79ef7a3cf", + "src/inferrsi/data/skill_library/repositories/vllm/diagnose-vllm-admission-pressure/references/source-notes.md": "84033fd514773f10c0bddd9914fabb980497ba2b308f226a922c3e1888413827", + "src/inferrsi/data/skill_library/repositories/vllm/references/evidence-map.json": "dd8bbab5d285ed3a1a19d2f06ad84e8c444251690d8f673def3e2681765ef6c1", + "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.json": "4e705a964697712e950e02bedc41320806a41311b7f4dcf32f923d60b1d2e919", + "src/inferrsi/data/skill_library/repositories/vllm/references/repo-provenance.md": "6e841c38c35ff2cd6ff8139cb1a6c8f75701647897f1089a94a5c62928609941", + "src/inferrsi/data/skill_library/repositories/vllm/references/repo-routing-metadata.json": "a20b54f682eb52311e445ac62b95d1935c001296f5cf23b7764cb04477d535c9", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/SKILL.md": "2a945b5fddacfb90d1cd7cb2b2f767067d641061423a65e1c19afecec7f3c06a", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/capability-boundaries.md": "c8081d03f04f7481080c46c7dcf728962729bc122f49f9502d50cd1e3d04bcf3", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/architecture-search-space/references/runtime-and-invariants.md": "afa1f967545021f36bc1b59dbf6aa2c367f4d5ece4f08dce755e2017cf50dbe5", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/SKILL.md": "495679e5b99341bf80daeaacd2d4c2699aba16bc16c447cbd8d2fe17f7559798", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/extend-validate/references/change-boundaries-and-validation.md": "f7fdd2d699f09d96fbf5ca9382f9b56113c870db2688d8e1118eb8163519fc4a", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/SKILL.md": "0ef4d2e3a96cbb277557e169a3fc8edf0f0504aad03259771b432a33ee008f78", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/commands-and-measurements.md": "76ab51083b14595c49fa1c255cb7e5ebb996f14930a51ab3cbe864cc334b5f80", + "src/inferrsi/data/skill_library/repositories/vllm/sub-skills/operation-performance/references/experiment-decisions.md": "242593c0d386e57104b57a09bc7db97c2bbbcdb905bb68a397cfe612e07fb083", + "src/inferrsi/data/skill_library/skill-index.json": "b5a9f05bd8179644f3fea79d7eb04fe4a8baad9e8d68aa8d0f1db213cf633317", + "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/SKILL.md": "485932cdf3b1fba5dc69be73e75d38f86a7b79442d26e82a296c963a7598d226", + "src/inferrsi/data/skill_library/workflows/cite-check-downstream-pr/metadata.json": "7a18f9dbb4894216802a44929b913990fc4bd8521cb60976d9435ee8f37e4265", + "src/inferrsi/data/skill_library/workflows/claim-contract/SKILL.md": "3af1dbbe81d91cb05fc2ef15a2ae25d7592a08558e1a5535f49c52f33a488c55", + "src/inferrsi/data/skill_library/workflows/claim-contract/metadata.json": "25c0137cf347f75d7ff8f8305d9ff5f055c6881ddf4b708f923c3176de16c153", + "src/inferrsi/data/skill_library/workflows/competitor-scan/SKILL.md": "0ede58ace65a9324a538b4816add1684f230ba536d4364fe5a5dd2c0bbed3776", + "src/inferrsi/data/skill_library/workflows/competitor-scan/metadata.json": "74b387e5df7b40cce9fd1ceee7ea1a26c258a4ed091144f2b366ef971895a2e8", + "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/SKILL.md": "ce983acfb7d7d3272775e7cf00c92e4ddadd97511b602575a76e5ab7ade01b56", + "src/inferrsi/data/skill_library/workflows/draft-acceptance-issue/metadata.json": "e97a665349f234d74baeaa20c6996bff990d0f45c40c0b7ac7283aa599cb951e", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/SKILL.md": "364beed2ca617e71f4f955bbd024740d411b94eeb9557d7e055d4463cc22caba", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/cross-repository-mechanisms.md": "28efa1b25f548763180b70d8d0ddebb7d4e7227df189f871fd233fe1d4881303", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/import.md": "c8a57db94768cc6cd4193932def75ba67ec453e56ee0e54f176fb82f1af53091", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/model-aware-preparation.md": "787c017aade68ee1133b979d636fa24062302cef6174c001171c963947ee8da4", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/overview.md": "7ed0a966239b9481b82dd349245cfbfe966c030fc916771c91fd0692464ffc56", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/repository-source-scope.md": "85202ca209171c50573c2fc209acb1e937d8211e1652e4ca9c6cc4bce04d03eb", + "src/inferrsi/data/skill_library/workflows/inference-engine-research/references/source-manifest.json": "0ce1678803d1ede6b485fb6b4b1c88aba909056ee6e3b5af691899ceb2056240", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/SKILL.md": "74cd863df2e80dc2bda1713d24207809173a3116f30325e3487e1e3517cfec1e", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/metadata.json": "073174963ca5c19c63d68843ccabdbf758f4c8b7ca61e3d7283affd6f88a753c", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/COMPARISON.md": "38587678a96581f5c2dc697d982127343565504bffa191352811dffcc165aa23", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/INTEGRATION_NOTES.md": "fbc4a96b6f6c477706eed83eaa4fcc29a14f881637629d6c004601ed049f2523", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/QA.md": "3f3785ec39a6938a5d1f2e874e99153fe92c0e666dd3e78a499c8b47d891346a", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/README.md": "221785e91c45588f82b3caab6d52f1ea1d102b16139c20a0802921e2789e6d28", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/UNREAD_LEADS.md": "6c1d87ebe491d27666703149eb05ad582c7e5aff77f0ac5c11a7b89c946f96ee", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/cases/methodology-counterexamples.md": "f9fc21753ea41eef77ca020d78d031e945e5dd835b5d09c5ff0e53749ceb2ca4", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/knowledge/evidence-routing.md": "406d5d74c3fef4c1a23d495492e21a084c3c46495a1a84b385703ed104358874", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/learning-to-commit.md": "9c2d1d738e896d05c00f53e6adf14c9df2501ecc3f0020e71b5a1c2192547086", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/protocol-details.md": "41298c6bde3a28f0712fa06a18bbc4ac59aea6d17cc9eb1318744f4121e3bab6", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/readset.md": "6f953d20d8958c68339c9ceeae348604d0fea3e9c6ccb9b64a63d3b210426a22", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/repo-to-skill.md": "5ebdac2a2bbbe6614f9d4096bc9895bd685acdf76414e2c3c1fb7b50d57c738d", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/skillopt.md": "148075daa67e19f4b4351710498438d17c279a9bce56b2d01cab26ff3be33e99", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/source-manifest.json": "716cd49ed01824ee9f6b6fc55ebf2b24fdd5d5a3e677f6e28e34a472616bbe25", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/references/trace2skill.md": "230d13b77ba819b5e3edee147523e8cad6bfce95afccc140fb4946d6c8747a2e", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/papers-20261015/updates/bounded-updates.md": "d02c5bd0109846d1e1987de074810321cb4af277cd3f66ab2ac268aadd5ee142", + "src/inferrsi/data/skill_library/workflows/npu-skill-factory/references/skill-factory-design-20261014.md": "fb8ce1ad83b7a6ddc657e89ee2eebb710873b213cfd85bdd6d823c8691b3cb9f", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/SKILL.md": "08f76e12f6f8e8893f03bcdc8152a96c0c99442e7faecbc2a67a18c632fcb243", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/metadata.json": "caf3ef265df71b3cbec31b6c4affb6ebe9278fba0626b23bc124825d2761e49a", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/LICENSE": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/LICENSE-MIT.md": "1df1cf519b9e5d8acbbc72820299d7aa95afa1acbfc1f43a15109ad1d59975f3", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/NOTICE.md": "3c3c36a40e1d65f3a6b20ec3a97319cb753e857f1d120f302cd828e107120f6d", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/README.md": "c5802bb4d71fbf95e40a6a5282c0afcba594ab5b34d40bec97874928d8376a00", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/decision-cases.md": "c2c1d3d75c4fcb515dd36ff84d8dc940c25d54c5757a29a80d1af88eb8b66115", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/mechanisms.md": "4fa09099d362a8088fdb0e6a767f72aae6218b92fe5fd75b6fb78c2bb2a318fb", + "src/inferrsi/data/skill_library/workflows/rsi-evidence/references/loopx-goal-governance/source-ledger.json": "60294951d702db924119300d89f4ed01d970a7036e214aa0d04c78f6842f92d9", + "src/inferrsi/evaluation.py": "482c40710d428206c59fe3696703248cd86c4b3c7d19ad41017c1369bd6d780e", + "src/inferrsi/evidence_paths.py": "cc46231e0b8334ef5742defdda20e0093e22b30c86af59df1fb3ae4b5caa70fa", + "src/inferrsi/execution.py": "f43d8cbb4d43a72a7ff8a10b29a5fb9cf4cb22af54dd35943283c843fba0da92", + "src/inferrsi/harness/README.md": "0fd486efbf4752e9e93b118102244cb332ceea203dfa3628cb70fbf816b9a110", + "src/inferrsi/harness/__init__.py": "1006e4713003f44bfb32b2e385a918e1652b9e9833801e4555674a71e57fb0d5", + "src/inferrsi/harness/evaluation.py": "20d0c52145600a5d6b1e513b61f2fffaa459b83db194876be40726bd4f343a59", + "src/inferrsi/harness/improvement.py": "3faa0f5928a707a98f35f45d1ca18e226c898941e6a07b3829eebeb53e70aeb8", + "src/inferrsi/harness/reporting.py": "38c58f02c44e8c8aeacf46527e142be0e3142a83e4b2693246d0ca76c0ddd7c3", + "src/inferrsi/harness/runner.py": "d290fc87710e77c8e6069e95bb0d347016a393707592d2eed98f07951f2d0865", + "src/inferrsi/harness/scenarios/__init__.py": "4e97913c806077f0cadfeb29d986d2fa75b55ab8766cb86e6e3e9a195a9a1f5a", + "src/inferrsi/harness/scenarios/external.py": "f3f8d4fc516ebce81c26d82248e83f79b299eed0fc17ce73e2844995aea9c3f9", + "src/inferrsi/harness/scenarios/media.py": "67d979a3ed6f13a8e3e1e40a304f78e5a3e6c2556b3aa8a11c41156745d7050a", + "src/inferrsi/harness/skill_eval_diagnostic.py": "e25cfee943052aaefbddf3f545fe6d56d95ab1edb9723cd9353feeabc954ea9a", + "src/inferrsi/harness/skill_eval_engineering.py": "2f18a1ab96acfc84319eba05e94c2a073fd0dca83e4f82ad98142ba2fecda039", + "src/inferrsi/harness/skill_evaluation.py": "ca141c3d1272591e24797d1ee31bed3aa37d9a009897c88c2ce0b087f3da11ac", + "src/inferrsi/infra/README.md": "8a3d41ba68cc655ef1788456e532125123902523f858b8932398995cc9127719", + "src/inferrsi/infra/__init__.py": "8779897733779266841b0f79b84b1dc43f0fa57c013855c906c809829ecfc77c", + "src/inferrsi/infra/artifacts.py": "bcf3adf325e34e10a4ecb4baf0af855ddca3d77eff76de5b66e493baaa3d7eaf", + "src/inferrsi/infra/campaign.py": "be505fba1592611bd84ceaca55afbac027205413a2639baefa0c3bb343230209", + "src/inferrsi/infra/commands.py": "1890f115b3a2c5c6832173bf60432ad595bd6bf1c69783ecd460c6f8961194f1", + "src/inferrsi/infra/environment.py": "23a6716089c542617761b5a3fc67dde0af9e483212fc003f8577757fbc4ed2db", + "src/inferrsi/infra/executors/__init__.py": "327c3cd5c821c832ff260a41dba0ee1e439cdabefc9edfb9cad1cd38d14bc87a", + "src/inferrsi/infra/executors/local.py": "37bcbca9f3261c8c58a8bf806c9da40720b18d0676c82e8ed24c3901ac9bbe77", + "src/inferrsi/infra/http.py": "f65d6c2b55b6a89694caaf6dcdf7d74f417e9ced103eca1b3f218e305c6b6cb2", + "src/inferrsi/infra/inspection.py": "d1c66deb0cced529bca996d2f88b60f51fa3417ff1cae365c650fec808946ecd", + "src/inferrsi/infra/platforms/__init__.py": "dac68eb6edbff995f8afe477396d81efd5db93128978ceedaea1babf0dd3db08", + "src/inferrsi/infra/platforms/cuda.py": "9e0e4276e1b2984605493b14b82fba3e86f6635204719c318302e624f66f8b6f", + "src/inferrsi/infra/platforms/mlx.py": "f3b81470db28bd6183ccca2e763c5e079686bba6cdfe6c08ebe2cdf64216f2e3", + "src/inferrsi/infra/processes.py": "3b96ebc7a50913e78e86ff3dcba74c8ee8d03d547d9591ca6488f390ac1e6189", + "src/inferrsi/infra/targets/__init__.py": "1f3ea75dded7c6cc967754c7b6d4f30ef00da8191a749938aabc3c6bb8f86eb3", + "src/inferrsi/infra/targets/git.py": "abb93f9332db336298102e28c06b9a7553597d4d03b957358602f60d61b1aa65", + "src/inferrsi/inspection.py": "afa44bf4cdfbb58a318c21f9b292ca142d1bbb22750e35a4cc8244da5350ebb3", + "src/inferrsi/knowledge_operations.py": "c1a5f0756cc1cdad66d7b766f68405b622d85d65311546a952ab77b66f0bde44", + "src/inferrsi/knowledge_store.py": "70dfca059b0a682579206d4376d191667934b3ca869ae847098c4c6964d86abe", + "src/inferrsi/model/README.md": "a8a4fefbc943c97b2fac6c4a05e751c90b1e79d46de8fb59f500e92f8ddd5e34", + "src/inferrsi/model/__init__.py": "3c1dba4f5159fd9fbc8d4d96e2a83c8441491c1b1dc7b42031a3fc1782ed9d9f", + "src/inferrsi/model/anthropic_agent.py": "746db742467fee153427cc9d7e8a01b640bc13e5b9cd431e935f596a8e19b072", + "src/inferrsi/model/protocols/__init__.py": "5a2f54483b164de04e309dbebb87a700b213682b0f44a50d89a752dd71a4ce47", + "src/inferrsi/model/protocols/image.py": "251a1f8f885da2cd8e9d3baba766c525afe293091606b4e8668c43ba2cb7cb5f", + "src/inferrsi/model/protocols/metrics.py": "5b32e296a746b6de30cb0c89f23f730296b0087884c1acc2b6127c1295d4db7a", + "src/inferrsi/model/request_guard.py": "c2f8a2f822f32785327a2138ee7b0609896b180819d9d1cbe851534aa6198628", + "src/inferrsi/model/validators/__init__.py": "1d4b6ffa968e283219ddcc8a09de9280238a8d65ae2dc8cd2b60889877a8eb0b", + "src/inferrsi/model/validators/flare.py": "3696b0609c772f2222b448833608565a033123d7e92b11fbc545dd356fec590f", + "src/inferrsi/model/validators/image.py": "79470b6b06a22d53ec6c40636a0b3d821a2e6ee2f7bb7aed3dac7241504add1a", + "src/inferrsi/reporting.py": "1f771890ab3e69b1aa9246b548a7825596bdf904f6c395dd218208084266853d", + "src/inferrsi/research.py": "9bfe59dd17b5476c18ce50d671711053c62800521c50d6b710e8140a2b4dca04", + "src/inferrsi/runspec.py": "dca0952425caf42095ef0c7e2feb8ef4740cc422ecb61e777f203962d9b87709", + "src/inferrsi/skill_catalog.py": "dbbe107a105e13b0c288f48ad80f9b05e108606af379d243dfab85cdfc99cfb5", + "src/inferrsi/summary_chart.py": "d29023d60f6be804c1b01c9eb73765d4824d6c3bb870fe22c99a24544bc1d8bc" + } +} diff --git a/evidence/indexer-255-token-v3-20261030/plan/synthetic-contracts.json b/evidence/indexer-255-token-v3-20261030/plan/synthetic-contracts.json new file mode 100644 index 000000000..a5232a0bb --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/plan/synthetic-contracts.json @@ -0,0 +1,282 @@ +{ + "status": "prospective before implementation; manually specified expected arithmetic, no candidate evaluated", + "baseline": "b3549ce1aea1556f76de1d2e14a4240474947b6b", + "cases": [ + { + "id": "token-dev01", + "cue": "ram", + "fields": [ + [ + 4, + "program analysis" + ], + [ + 1, + "RAM pressure" + ] + ], + "expected_whole_token_score": 1 + }, + { + "id": "token-dev02", + "cue": "io", + "fields": [ + [ + 4, + "optimization" + ], + [ + 3, + "I/O telemetry" + ], + [ + 1, + "io tracing" + ] + ], + "expected_whole_token_score": 1 + }, + { + "id": "token-dev03", + "cue": "GPU gpu", + "fields": [ + [ + 4, + "gpu GPU" + ], + [ + 1, + "GPU" + ] + ], + "expected_whole_token_score": 5 + }, + { + "id": "token-dev04", + "cue": "cpu.stat.local", + "fields": [ + [ + 4, + "cpu stat local" + ], + [ + 1, + "cpu.stat.local" + ] + ], + "expected_whole_token_score": 15 + }, + { + "id": "token-dev05", + "cue": "read_cache", + "fields": [ + [ + 4, + "read cache" + ], + [ + 1, + "read_cache" + ] + ], + "expected_whole_token_score": 1 + }, + { + "id": "token-dev06", + "cue": "checkpoint", + "fields": [ + [ + 4, + "checkpoints" + ], + [ + 1, + "checkpoint" + ] + ], + "expected_whole_token_score": 1 + }, + { + "id": "token-dev07", + "cue": "\u6570\u636e", + "fields": [ + [ + 4, + "\u6570\u636e\u5904\u7406" + ], + [ + 1, + "\u6570\u636e" + ] + ], + "expected_whole_token_score": 1 + }, + { + "id": "token-dev08", + "cue": "orchid", + "fields": [ + [ + 4, + "memory allocator" + ], + [ + 1, + "stream ownership" + ] + ], + "expected_whole_token_score": 0 + }, + { + "id": "string-dev09", + "cue": "whether next", + "fields": [ + [ + 1, + "prefix\nwhether\tnext" + ] + ], + "expected_whole_token_score": 2 + }, + { + "id": "string-dev10", + "cue": "whether next", + "fields": [ + [ + 1, + "prefix whether next" + ] + ], + "expected_whole_token_score": 2 + }, + { + "id": "string-dev11", + "cue": "quote beta", + "fields": [ + [ + 3, + "\"quote\" \\ beta" + ] + ], + "expected_whole_token_score": 6 + }, + { + "id": "string-dev12", + "cue": "input_contract source", + "fields": [ + [ + 2, + { + "input_contract": "source" + } + ] + ], + "expected_whole_token_score": 2 + }, + { + "id": "string-dev13", + "cue": "gpu cpu ram", + "fields": [ + [ + 2, + [ + "GPU", + { + "x": [ + "CPU", + "RAM", + "gpu" + ] + } + ] + ] + ], + "expected_whole_token_score": 6 + }, + { + "id": "string-dev14", + "cue": "null true 42", + "fields": [ + [ + 2, + [ + null, + true, + 42 + ] + ] + ], + "expected_whole_token_score": 0 + }, + { + "id": "string-dev15", + "cue": "null true 42", + "fields": [ + [ + 2, + [ + "null", + "true", + "42" + ] + ] + ], + "expected_whole_token_score": 6 + }, + { + "id": "string-dev16", + "cue": "download", + "fields": [ + [ + 3, + "downloads" + ], + [ + 1, + "download" + ] + ], + "expected_whole_token_score": 1 + }, + { + "id": "string-dev17", + "cue": "ram", + "fields": [ + [ + 4, + "program" + ], + [ + 1, + { + "ram": "dram" + } + ] + ], + "expected_whole_token_score": 0 + }, + { + "id": "string-dev18", + "cue": "read_cache", + "fields": [ + [ + 1, + "read_cache" + ], + [ + 4, + "read cache" + ] + ], + "expected_whole_token_score": 1 + } + ], + "unchanged_contracts": [ + "default mode byte parity", + "explicit ID order/membership/revision", + "query substring AND matching", + "filters before ranking", + "mandatory-resource byte/hash/revision and budget omission", + "legacy optional-resource scorer", + "qualified-ID deterministic tie order" + ] +} diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-default.json b/evidence/indexer-255-token-v3-20261030/results/baseline-default.json new file mode 100644 index 000000000..c776ce277 --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/results/baseline-default.json @@ -0,0 +1,53 @@ +[ + { + "options": {}, + "sha256": "cb428aabd177dd08b34caff4b247d4b11cfaa17308a6186f16f33e53b21f8ea6", + "bytes": 506405 + }, + { + "options": { + "relevance": "cgroupv2", + "context_bytes": 262144 + }, + "sha256": "c7b5eaee511ee846327c77ebf2891f1fa49f7472a4be9b2a648e6c326088e6bc", + "bytes": 163686 + }, + { + "options": { + "query": "stat.loc", + "relevance": "cgroupv2" + }, + "sha256": "b4ad59650fdf7f163efd0b172ede4050979245bfeff17aa37bb792db90ae5570", + "bytes": 153255 + }, + { + "options": { + "relevance": "gpu", + "backend": "ascend" + }, + "sha256": "ee60651afda9a7d06e16b480476ae504147d4fe768fa6a644ee50b02a60531d7", + "bytes": 1547320 + }, + { + "options": { + "ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "relevance": "ram", + "context_bytes": 262144 + }, + "sha256": "c8c9c2fd165f74998536a45ed2cd339c5a71abe1fce808de4ac0d98fdf3dc236", + "bytes": 163682 + }, + { + "options": { + "ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "relevance": "ram", + "context_bytes": 1 + }, + "sha256": "fc5d0f6b100d652e37fc5b59568ad8bcaca6d1b8e6dba577107638dac5830732", + "bytes": 153583 + } +] diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev01.json.gz b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev01.json.gz new file mode 100644 index 000000000..8f2c20b91 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev01.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev02.json.gz b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev02.json.gz new file mode 100644 index 000000000..d9e807c77 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev02.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev03.json.gz b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev03.json.gz new file mode 100644 index 000000000..b025e70b0 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev03.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev04.json.gz b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev04.json.gz new file mode 100644 index 000000000..c1b8e470f Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev04.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev05.json.gz b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev05.json.gz new file mode 100644 index 000000000..12a4a7e0b Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev05.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev06.json.gz b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev06.json.gz new file mode 100644 index 000000000..c7388b74a Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/baseline-freshdev06.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-freshnegative01.json.gz b/evidence/indexer-255-token-v3-20261030/results/baseline-freshnegative01.json.gz new file mode 100644 index 000000000..a8e527c8f Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/baseline-freshnegative01.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-freshnegative02.json.gz b/evidence/indexer-255-token-v3-20261030/results/baseline-freshnegative02.json.gz new file mode 100644 index 000000000..b2cc139a1 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/baseline-freshnegative02.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/baseline-summary.json b/evidence/indexer-255-token-v3-20261030/results/baseline-summary.json new file mode 100644 index 000000000..b51f01a16 --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/results/baseline-summary.json @@ -0,0 +1,152 @@ +[ + { + "id": "freshdev01", + "rank": null, + "ids": [ + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.audit_sd_cpp_residency", + "inferrsi.common.skills.npu_skill_factory", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.ascend.skills.cann_attribution_overlap" + ], + "complete_relevant": [], + "output_bytes": 1023627, + "context_bytes": 209844, + "context_omissions": [], + "sha256": "a2e2d4cb94d5e14bb3ad251d1b5a25d088dc5ab29918e36f5fe42d6f86e71d5c" + }, + { + "id": "freshdev02", + "rank": 1, + "ids": [ + "inferrsi.common.skills.choose_parallel_training_plan", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.training_input_pipeline", + "inferrsi.common.skills.training_resource_budget", + "inferrsi.cuda.skills.audit_flare_session_lifecycle" + ], + "complete_relevant": [ + "inferrsi.common.skills.choose_parallel_training_plan" + ], + "output_bytes": 995468, + "context_bytes": 192276, + "context_omissions": [], + "sha256": "052429227fab7d925a9846eca3bb567d948008e17566ca62827a72b2a1da194a" + }, + { + "id": "freshdev03", + "rank": 1, + "ids": [ + "inferrsi.common.skills.training_input_pipeline", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.kv_tier_management", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.common.skills.repository_evolution_analysis" + ], + "complete_relevant": [ + "inferrsi.common.skills.training_input_pipeline" + ], + "output_bytes": 936745, + "context_bytes": 141145, + "context_omissions": [], + "sha256": "459e5bff3a0cd751c64d0459e3754fc06548bf878d33cf8e7af98aab99da5275" + }, + { + "id": "freshdev04", + "rank": 2, + "ids": [ + "inferrsi.common.skills.npu_skill_factory", + "inferrsi.common.skills.skill_evolution_playbook", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.common.skills.weight_loading_optimization" + ], + "complete_relevant": [ + "inferrsi.common.skills.skill_evolution_playbook" + ], + "output_bytes": 991347, + "context_bytes": 186469, + "context_omissions": [], + "sha256": "fa56817e4ef5c18d55d3cd981ebd8067599e8e6d48a68fad0d3a2b556d926069" + }, + { + "id": "freshdev05", + "rank": 1, + "ids": [ + "inferrsi.cuda.skills.reason_hopper_kernels", + "inferrsi.ascend.skills.cann_attribution_overlap", + "inferrsi.ascend.skills.ascend_distributed_parallel", + "inferrsi.common.skills.draft_acceptance_issue", + "inferrsi.common.skills.weight_loading_optimization" + ], + "complete_relevant": [ + "inferrsi.cuda.skills.reason_hopper_kernels" + ], + "output_bytes": 1024332, + "context_bytes": 217296, + "context_omissions": [ + { + "ref": "package:inferrsi.common.skills.weight_loading_optimization", + "reason": "context_limit", + "required_bytes": 47137, + "available_bytes": 44848 + } + ], + "sha256": "e5af4a11f36aeb6754e554419cdda3226bd6d4d90e377a67a2fff6f815923adb" + }, + { + "id": "freshdev06", + "rank": null, + "ids": [ + "inferrsi.ascend.skills.ascend_operator_tuning", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.ascend.skills.ascend_operator_profiling", + "inferrsi.common.skills.storage_io_measurement", + "inferrsi.ascend.skills.ascend_video_diffusion_optimization" + ], + "complete_relevant": [], + "output_bytes": 908046, + "context_bytes": 114329, + "context_omissions": [], + "sha256": "48acc9062fdb58f0e04bf83fd53dd43c2b2fc12b1690fd3f2cfc0be284d82b08" + }, + { + "id": "freshnegative01", + "rank": null, + "ids": [ + "inferrsi.common.skills.dense_performance_evidence", + "inferrsi.common.skills.optimization_search_policy", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.ascend.skills.ascend_optimization_casebook" + ], + "complete_relevant": [], + "output_bytes": 938511, + "context_bytes": 140102, + "context_omissions": [ + { + "ref": "package:inferrsi.common.skills.performance_ceiling_analysis", + "reason": "context_limit", + "required_bytes": 170782, + "available_bytes": 131214 + } + ], + "sha256": "dcaa7d4e51b2bd26d89987810f2b65f517e4dc1362caa9d672ed8dc0fcdc2c25" + }, + { + "id": "freshnegative02", + "rank": null, + "ids": [ + "inferrsi.common.skills.npu_skill_factory", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.ascend.skills.ascend_operator_tuning", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.common.skills.competitor_scan" + ], + "complete_relevant": [], + "output_bytes": 968466, + "context_bytes": 163968, + "context_omissions": [], + "sha256": "dba7c45a509cf178f694a28f3e3294305a4878395b6c886454c5fddd13cb129d" + } +] diff --git a/evidence/indexer-255-token-v3-20261030/results/candidate-default.json b/evidence/indexer-255-token-v3-20261030/results/candidate-default.json new file mode 100644 index 000000000..c776ce277 --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/results/candidate-default.json @@ -0,0 +1,53 @@ +[ + { + "options": {}, + "sha256": "cb428aabd177dd08b34caff4b247d4b11cfaa17308a6186f16f33e53b21f8ea6", + "bytes": 506405 + }, + { + "options": { + "relevance": "cgroupv2", + "context_bytes": 262144 + }, + "sha256": "c7b5eaee511ee846327c77ebf2891f1fa49f7472a4be9b2a648e6c326088e6bc", + "bytes": 163686 + }, + { + "options": { + "query": "stat.loc", + "relevance": "cgroupv2" + }, + "sha256": "b4ad59650fdf7f163efd0b172ede4050979245bfeff17aa37bb792db90ae5570", + "bytes": 153255 + }, + { + "options": { + "relevance": "gpu", + "backend": "ascend" + }, + "sha256": "ee60651afda9a7d06e16b480476ae504147d4fe768fa6a644ee50b02a60531d7", + "bytes": 1547320 + }, + { + "options": { + "ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "relevance": "ram", + "context_bytes": 262144 + }, + "sha256": "c8c9c2fd165f74998536a45ed2cd339c5a71abe1fce808de4ac0d98fdf3dc236", + "bytes": 163682 + }, + { + "options": { + "ids": [ + "inferrsi.common.skills.cgroup_resource_controls" + ], + "relevance": "ram", + "context_bytes": 1 + }, + "sha256": "fc5d0f6b100d652e37fc5b59568ad8bcaca6d1b8e6dba577107638dac5830732", + "bytes": 153583 + } +] diff --git a/evidence/indexer-255-token-v3-20261030/results/compatibility-tests.txt b/evidence/indexer-255-token-v3-20261030/results/compatibility-tests.txt new file mode 100644 index 000000000..ac614edd5 --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/results/compatibility-tests.txt @@ -0,0 +1,55 @@ +................................................... [100%] +==================================== PASSES ==================================== +=========================== short test summary info ============================ +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[token-dev01] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[token-dev02] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[token-dev03] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[token-dev04] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[token-dev05] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[token-dev06] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[token-dev07] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[token-dev08] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev09] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev10] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev11] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev12] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev13] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev14] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev15] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev16] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev17] +PASSED tests/test_collect_token_selection.py::test_frozen_token_contract[string-dev18] +PASSED tests/test_collect_token_selection.py::test_default_and_explicit_default_identical +PASSED tests/test_collect_token_selection.py::test_opt_in_validation +PASSED tests/test_collect_token_selection.py::test_explicit_selection_resources_and_budget_unchanged +PASSED tests/test_collect_token_selection.py::test_token_operational_controls_and_substring_query +PASSED tests/test_collect_token_selection.py::test_token_ties_use_qualified_identity_order +PASSED tests/test_collect_token_selection.py::test_cli_token_method +PASSED tests/test_collect_selection.py::test_frozen_offline_pilot[cgroupv2--inferrsi.common.skills.cgroup_resource_controls-True] +PASSED tests/test_collect_selection.py::test_frozen_offline_pilot[MiniMax-H3 AdaLN SSD--inferrsi.common.skills.audit_h3_runtime_residency-False] +PASSED tests/test_collect_selection.py::test_frozen_offline_pilot[P2pNccl--inferrsi.cuda.skills.audit_afd_p2p_exchanges-True] +PASSED tests/test_collect_selection.py::test_frozen_offline_pilot[DLRover replica recovery--inferrsi.common.skills.checkpoint_recovery_consistency-False] +PASSED tests/test_collect_selection.py::test_frozen_offline_pilot[cgroupv2-cpu.stat.local-inferrsi.common.skills.cgroup_resource_controls-False] +PASSED tests/test_collect_selection.py::test_no_match_and_query_remains_token_and_substring_matching +PASSED tests/test_collect_selection.py::test_existing_item_payload_and_material_accounting_are_unchanged +PASSED tests/test_collect_selection.py::test_explicit_membership_order_aliases_and_revisions_are_preserved +PASSED tests/test_collect_selection.py::test_filters_stay_package_only_and_precede_ranking +PASSED tests/test_collect_selection.py::test_weighted_presence_and_identity_ties_are_deterministic +PASSED tests/test_collect_selection.py::test_context_keeps_complete_core_prerequisites_provenance_and_exact_bytes +PASSED tests/test_collect_selection.py::test_both_budgets_skip_oversized_items_and_fit_later_smaller_items +PASSED tests/test_collect_selection.py::test_source_drift_and_cross_owner_mapping_never_fall_back +PASSED tests/test_collect_selection.py::test_store_delegates_but_copied_package_metadata_cannot_gain_context +PASSED tests/test_collect_selection.py::test_h3_readsets_are_mandatory_even_without_lexical_overlap +PASSED tests/test_collect_selection.py::test_parent_owner_license_is_mandatory_and_budgeted_for_cross_skill_references +PASSED tests/test_collect_selection.py::test_retained_legal_resources_ignore_relevance_and_use_atomic_budget[inferrsi.cuda.skills.audit_afd_p2p_exchanges-P2pNccl-data/skill_library/repositories/afd-plugin/audit-afd-p2p-exchanges/references/afd-plugin-LICENSE.md] +PASSED tests/test_collect_selection.py::test_retained_legal_resources_ignore_relevance_and_use_atomic_budget[inferrsi.common.skills.audit_h3_runtime_residency-orchid-data/skill_library/repositories/h3-c/audit-h3-runtime-residency/references/THIRD_PARTY_NOTICES.md] +PASSED tests/test_collect_selection.py::test_missing_declared_notice_is_explicitly_unavailable +PASSED tests/test_collect_selection.py::test_optimized_request_preserves_identity_collisions +PASSED tests/test_collect_selection.py::test_invalid_opt_ins_fail_at_request_boundary[options0] +PASSED tests/test_collect_selection.py::test_invalid_opt_ins_fail_at_request_boundary[options1] +PASSED tests/test_collect_selection.py::test_invalid_opt_ins_fail_at_request_boundary[options2] +PASSED tests/test_collect_selection.py::test_invalid_opt_ins_fail_at_request_boundary[options3] +PASSED tests/test_collect_selection.py::test_invalid_opt_ins_fail_at_request_boundary[options4] +PASSED tests/test_collect_selection.py::test_invalid_opt_ins_fail_at_request_boundary[options5] +PASSED tests/test_collect_selection.py::test_cli_exposes_context_in_json_and_text +51 passed in 25.39s diff --git a/evidence/indexer-255-token-v3-20261030/results/independent-review.json b/evidence/indexer-255-token-v3-20261030/results/independent-review.json new file mode 100644 index 000000000..38aa8d2de --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/results/independent-review.json @@ -0,0 +1,57 @@ +{ + "baseline": { + "ranks": [ + null, + 1, + 1, + 2, + 1, + null + ], + "hit": 4, + "mrr": 0.5833333333333334, + "complete_relevant": 4, + "audited_contexts": 38, + "raw_gzip_sha256": { + "baseline-freshdev01.json.gz": "9683515a0a90125edbc2a494ff9d97e03654eb7700695a58b61f17fb5414aafd", + "baseline-freshdev02.json.gz": "3ac8771f3e40e14eb9da47cd1284c683491f8ee4968a89770395d3b9c90a6c49", + "baseline-freshdev03.json.gz": "24df4affd3879986cb4089c84243d8f3f7afbddf1038e1d502de751029b832b9", + "baseline-freshdev04.json.gz": "9fb82c97f1ac3f67abef7706f50118815cf35e6b14eab82ab9be6b05cd53b3ff", + "baseline-freshdev05.json.gz": "c2211daf1ffabfff0e90c3bad2e6e6f5a84cf53c9eb0b2b14cdb9380f0c5a543", + "baseline-freshdev06.json.gz": "4bec8b9ce9077248e8f002bb13567664df71168ea37688abfed372ca5a24307d", + "baseline-freshnegative01.json.gz": "7b70948a577c3aa8f430718309df9ce8e8a4ab8ef7054472f74dd173fe725bab", + "baseline-freshnegative02.json.gz": "4b96926e945b8b31f968e786e6e5a618c3101922a7e5f829e401833fed09f8c6" + } + }, + "token": { + "ranks": [ + 5, + 1, + 2, + 2, + 1, + null + ], + "hit": 5, + "mrr": 0.5333333333333333, + "complete_relevant": 5, + "audited_contexts": 38, + "raw_gzip_sha256": { + "token-freshdev01.json.gz": "3f4a5484715035baeadd92a0e64f7d9402dafff9eccf84a3dff517c17a390bb5", + "token-freshdev02.json.gz": "428c80858059622c5fe654a5ec41304510186c40aeda461917b23d2a2fef72ee", + "token-freshdev03.json.gz": "df04a606910247ad3303ca0dc6139ee8111bfd2dd7a9b4568387b679d7271899", + "token-freshdev04.json.gz": "38dab17bc5e0ee8b848c45f533fb94090863844794b75908da94da2871a2dfb4", + "token-freshdev05.json.gz": "5fa0d8aaade3ccf76b860786fdb0957d77c585e950c7d1b34fa906bb82beb900", + "token-freshdev06.json.gz": "67844db80222a1cb5827bc9e16ccade2643ece100f9700b84f75bf227780cbcb", + "token-freshnegative01.json.gz": "43e35e13109b1b18514a43e72a44ff447b965fe9dd6405045eaf2b5453f676a3", + "token-freshnegative02.json.gz": "9a86fbcd273dbfd50b8d232684145457cad9083988a70b530f9bb2789147e28e" + } + }, + "candidate": "96788d26b57240e53c26fe5446b2062a62ee6d0e", + "baseline_commit": "b3549ce1aea1556f76de1d2e14a4240474947b6b", + "source_and_runner_freeze_verified": true, + "default_capture_parity": true, + "heldout": "NOT_EXECUTED; content not accessed in this review", + "scope": "Supplemental retained-output read/required-resource audit; no collect/retrieval rerun, no test rerun; original 51-pass log inspected separately", + "decision": "FAIL development MRR gate; stop before heldout, no runtime promotion or255closure" +} diff --git a/evidence/indexer-255-token-v3-20261030/results/report.md b/evidence/indexer-255-token-v3-20261030/results/report.md new file mode 100644 index 000000000..78c47e90f --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/results/report.md @@ -0,0 +1,50 @@ +# Issue #255: structured string-value tokenization v3 — FAIL + +## Decision +The single fixed v3 candidate fails the prospective development MRR gate. Stop: do not promote it, do not run heldout, and do not close #255. No tuning, additional candidate, morphology, model calls or timing campaign was performed. Historical v2 FAIL remains unchanged. + +| Partial-positive DEV metric | Production baseline | Historical v2 (not rerun) | v3 | +|---|---:|---:|---:| +| Hit@5 | 4/6 | 5/6 | 5/6 | +| MRR@5 | 0.583333 | 0.541667 | 0.533333 | +| Complete relevant deliveries | 4 | 5 | 5 | + +These six DEV cases are reused training diagnostics, not fresh validation. The six sealed heldout cases were not exposed to the implementer and were not evaluated; their original author/source reviewers necessarily accessed labels for preparation. No heldout content is included here. + +## Frozen identity +- Production baseline: `b3549ce1aea1556f76de1d2e14a4240474947b6b`; source matches all813 entries in the prospective source manifest and is byte-identical to prior1147 production src. +- Candidate: `96788d26b57240e53c26fe5446b2062a62ee6d0e`; tree `5ae435fb5793b19f0de96ce74dd6ded9db3f3292`. +- Protocol SHA256: `2adf6a226b4c3763e52dec159aaae9ec4ea0824f2b0b0bd37a92799ef06bf94d`. +- Code and runner were frozen before semantic DEV evaluation: see pre-semantic-freeze.json. That receipt's semantic_runs_started=false describes the freeze instant, not final run status. +- Frozen evaluator SHA256: `47570bda5dd642c9a7170dcd24680806a007ecc88432569af9afa4867dbc7db8`. + +## Single change tested +Explicit opt-in `weighted-token-v3` recursively unions tokens from scalar string VALUES within each weighted field. Dictionary keys and nonstring values contribute nothing; list/tuple/dict-values recurse. Strings are tokenized separately, not joined. Existing Unicode words regex/casefold and weights4/3/2/1 are unchanged. No stemming, synonyms, query/corpus/metadata changes. The optional-resource lexical scorer and default weighted-lexical-v1 remain unchanged. Key removal is an explicit policy consequence, not merely a newline repair. + +## Per-case diagnostics +Partial positive labels do not judge every returned item. The two domain-negative cases are nongating and their unlabelled returns are not false-positive labels. + +| Case | Production rank | Historical v2 rank | v3 rank | Relevant complete A/B | Context bytes A/B | Output bytes A/B | +|---|---:|---:|---:|---:|---:|---:| +| freshdev01 | miss/unlabelled | 4 | 5 | 0/1 | 209844/186748 | 1023627/997003 | +| freshdev02 | 1 | 1 | 1 | 1/1 | 192276/151271 | 995468/948177 | +| freshdev03 | 1 | 2 | 2 | 1/1 | 141145/183819 | 936745/984882 | +| freshdev04 | 2 | 2 | 2 | 1/1 | 186469/169204 | 991347/973224 | +| freshdev05 | 1 | 1 | 1 | 1/1 | 217296/254712 | 1024332/1064339 | +| freshdev06 | miss/unlabelled | miss/unlabelled | miss/unlabelled | 0/0 | 114329/257046 | 908046/1062990 | +| freshnegative01 | miss/unlabelled | miss/unlabelled | miss/unlabelled | 0/0 | 140102/166489 | 938511/965135 | +| freshnegative02 | miss/unlabelled | miss/unlabelled | miss/unlabelled | 0/0 | 163968/175888 | 968466/986613 | + +Freshdev01 moves from production miss to rank5 (v2 rank4), while freshdev03 remains rank2 rather than production rank1. The rank5 gain contributes0.2 but rank1→2 costs0.5 in reciprocal-rank sum. This experiment does not establish general quality improvement or isolate which token-policy component caused each observed rank shift. + +## Verification +- 51 focused tests passed in25.39s, including18 frozen synthetic contracts and existing/new compatibility tests. Ruff source/new-test check passed; not a full-suite claim. +- Six baseline/candidate default serialized captures match exactly, including default, relevance/context, substring query, filter, explicit ID and undersized budget scenarios. +- Frozen evaluator verifies complete indexed_skill required-resource sets and exact emitted bytes/hash with pinned revision reads for every emitted context. Relevant complete counts4→5 remain separate from retrieval misses/budget omissions. Full raw gzip outputs retain omissions and control metadata. +- Independent supplemental audit (`independent-review.json`) verifies38 emitted contexts per arm, complete required paths/bytes/hash/revision and context-budget accounting, all16 raw gzip/summary hashes, default parity and frozen code/runner identity. This post-run audit did not execute retrieval or heldout. +- No runtime code changed after pre-semantic freeze. Historical v2 scores are a separate diagnostic column from saved outputs, not pooled or rerun. Gate comparator remains production baseline. + +## Replay / limits +Candidate.patch applies to pinned baseline; candidate.bundle retains exact candidate with baseline prerequisite. Use an existing Python/PyYAML environment, without new installation or paid calls. Set PREP_ROOT to inputs/, create OUTPUT_DIR with `mkdir -p "$OUTPUT_DIR"`, and execute frozen evaluate_development.py with BASELINE_ROOT PREP_ROOT OUTPUT_DIR baseline or CANDIDATE_ROOT PREP_ROOT OUTPUT_DIR token. Use a new output directory, not frozen results. compatibility_capture.py ROOT emits the retained parity capture. The frozen synthetic fixture lives in the retained candidate patch. Gzip results are lossless with mtime0. + +Small diagnostic partial labels are not complete Recall/precision, population accuracy, downstream agent benefit or hardware efficacy. No speed/nonregression conclusion: this trial is untimed. Gate FAIL cannot be repaired by weakening the comparator or evaluating/tuning on heldout. Any further hypothesis requires a separate prospective decision. diff --git a/evidence/indexer-255-token-v3-20261030/results/token-freshdev01.json.gz b/evidence/indexer-255-token-v3-20261030/results/token-freshdev01.json.gz new file mode 100644 index 000000000..2f4193db9 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/token-freshdev01.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/token-freshdev02.json.gz b/evidence/indexer-255-token-v3-20261030/results/token-freshdev02.json.gz new file mode 100644 index 000000000..df5a2fdd5 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/token-freshdev02.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/token-freshdev03.json.gz b/evidence/indexer-255-token-v3-20261030/results/token-freshdev03.json.gz new file mode 100644 index 000000000..efc6de4e3 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/token-freshdev03.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/token-freshdev04.json.gz b/evidence/indexer-255-token-v3-20261030/results/token-freshdev04.json.gz new file mode 100644 index 000000000..6c3132e57 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/token-freshdev04.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/token-freshdev05.json.gz b/evidence/indexer-255-token-v3-20261030/results/token-freshdev05.json.gz new file mode 100644 index 000000000..050abe31a Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/token-freshdev05.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/token-freshdev06.json.gz b/evidence/indexer-255-token-v3-20261030/results/token-freshdev06.json.gz new file mode 100644 index 000000000..350121f2a Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/token-freshdev06.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/token-freshnegative01.json.gz b/evidence/indexer-255-token-v3-20261030/results/token-freshnegative01.json.gz new file mode 100644 index 000000000..ac22a1d90 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/token-freshnegative01.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/token-freshnegative02.json.gz b/evidence/indexer-255-token-v3-20261030/results/token-freshnegative02.json.gz new file mode 100644 index 000000000..5ff78cf20 Binary files /dev/null and b/evidence/indexer-255-token-v3-20261030/results/token-freshnegative02.json.gz differ diff --git a/evidence/indexer-255-token-v3-20261030/results/token-summary.json b/evidence/indexer-255-token-v3-20261030/results/token-summary.json new file mode 100644 index 000000000..02ece076d --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/results/token-summary.json @@ -0,0 +1,153 @@ +[ + { + "id": "freshdev01", + "rank": 5, + "ids": [ + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.cuda.skills.diagnose_attention_bottlenecks", + "inferrsi.common.skills.training_input_pipeline", + "inferrsi.common.skills.npu_skill_factory", + "inferrsi.common.skills.assess_split_backward_schedule" + ], + "complete_relevant": [ + "inferrsi.common.skills.assess_split_backward_schedule" + ], + "output_bytes": 997003, + "context_bytes": 186748, + "context_omissions": [], + "sha256": "de9441c598ecdb0ca302b2e255b95b470ed4cbc9cc68964bbdce98976e508f56" + }, + { + "id": "freshdev02", + "rank": 1, + "ids": [ + "inferrsi.common.skills.choose_parallel_training_plan", + "inferrsi.common.skills.training_input_pipeline", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.training_resource_budget", + "inferrsi.ascend.skills.cann_attribution_overlap" + ], + "complete_relevant": [ + "inferrsi.common.skills.choose_parallel_training_plan" + ], + "output_bytes": 948177, + "context_bytes": 151271, + "context_omissions": [], + "sha256": "8cde0c29fa64c5491419a197ec9fb3d1b9c171164073577ee70083f9a074d763" + }, + { + "id": "freshdev03", + "rank": 2, + "ids": [ + "inferrsi.common.skills.kv_tier_management", + "inferrsi.common.skills.training_input_pipeline", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.common.skills.budget_speculative_verification", + "inferrsi.common.skills.local_dream" + ], + "complete_relevant": [ + "inferrsi.common.skills.training_input_pipeline" + ], + "output_bytes": 984882, + "context_bytes": 183819, + "context_omissions": [], + "sha256": "5e60171ec817882a431b9a1da97af692818cc9811ca3c2138c68abea9c64a85d" + }, + { + "id": "freshdev04", + "rank": 2, + "ids": [ + "inferrsi.common.skills.npu_skill_factory", + "inferrsi.common.skills.skill_evolution_playbook", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.common.skills.storage_io_measurement", + "inferrsi.common.skills.budget_speculative_verification" + ], + "complete_relevant": [ + "inferrsi.common.skills.skill_evolution_playbook" + ], + "output_bytes": 973224, + "context_bytes": 169204, + "context_omissions": [], + "sha256": "09c21d3fa2a1d50c7026f63b29b085c4c8a57b7589bdc242255986832caa484d" + }, + { + "id": "freshdev05", + "rank": 1, + "ids": [ + "inferrsi.cuda.skills.reason_hopper_kernels", + "inferrsi.common.skills.kernel_writing_guide", + "inferrsi.common.skills.audit_llama_runtime_state", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.ascend.skills.ascend_optimization_casebook" + ], + "complete_relevant": [ + "inferrsi.cuda.skills.reason_hopper_kernels" + ], + "output_bytes": 1064339, + "context_bytes": 254712, + "context_omissions": [ + { + "ref": "package:inferrsi.common.skills.kernel_writing_guide", + "reason": "context_limit", + "required_bytes": 126421, + "available_bytes": 79141 + }, + { + "ref": "package:inferrsi.ascend.skills.ascend_optimization_casebook", + "reason": "context_limit", + "required_bytes": 9172, + "available_bytes": 7432 + } + ], + "sha256": "132618c5be9924d1adaef906f81d74007e166d3f7eacc29e18d82a565d517b45" + }, + { + "id": "freshdev06", + "rank": null, + "ids": [ + "inferrsi.ascend.skills.ascend_operator_tuning", + "inferrsi.common.skills.performance_ceiling_analysis", + "inferrsi.common.skills.storage_io_measurement", + "inferrsi.common.skills.weight_loading_optimization", + "inferrsi.ascend.skills.cann_attribution_overlap" + ], + "complete_relevant": [], + "output_bytes": 1062990, + "context_bytes": 257046, + "context_omissions": [], + "sha256": "ad2b3c077ec5dfafe860f64335ff436be135ea7b91a10562abf65edc6b23c045" + }, + { + "id": "freshnegative01", + "rank": null, + "ids": [ + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.common.skills.optimization_search_policy", + "inferrsi.ascend.skills.ascend_optimization_casebook", + "inferrsi.common.skills.bound_windowed_attention_replay", + "inferrsi.common.skills.skill_evolution_playbook" + ], + "complete_relevant": [], + "output_bytes": 965135, + "context_bytes": 166489, + "context_omissions": [], + "sha256": "4f88030f316e92a878da4480d8d2ce900d02b530cdfce1fdecc1e2db7e6396f4" + }, + { + "id": "freshnegative02", + "rank": null, + "ids": [ + "inferrsi.common.skills.npu_skill_factory", + "inferrsi.common.skills.repository_evolution_analysis", + "inferrsi.common.skills.budget_speculative_verification", + "inferrsi.common.skills.inference_optimization_casebook", + "inferrsi.common.skills.competitor_scan" + ], + "complete_relevant": [], + "output_bytes": 986613, + "context_bytes": 175888, + "context_omissions": [], + "sha256": "5ac8d68e05640b4e0fc0506d90f85bfc9ba709ca0096b0fd2738411867427f5f" + } +] diff --git a/evidence/indexer-255-token-v3-20261030/validator/compatibility_capture.py b/evidence/indexer-255-token-v3-20261030/validator/compatibility_capture.py new file mode 100644 index 000000000..fb4bdc3cc --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/validator/compatibility_capture.py @@ -0,0 +1,10 @@ +import hashlib,json,sys +from pathlib import Path +root=Path(sys.argv[1]).resolve();sys.path.insert(0,str(root/'src')) +from inferrsi.knowledge_operations import KnowledgeReader +reader=KnowledgeReader(); cg='inferrsi.common.skills.cgroup_resource_controls' +rows=[] +for opts in ({},{'relevance':'cgroupv2','context_bytes':262144},{'query':'stat.loc','relevance':'cgroupv2'},{'relevance':'gpu','backend':'ascend'},{'ids':[cg],'relevance':'ram','context_bytes':262144},{'ids':[cg],'relevance':'ram','context_bytes':1}): + raw=json.dumps(reader.collect(**opts),sort_keys=True,ensure_ascii=False).encode() + rows.append({'options':opts,'sha256':hashlib.sha256(raw).hexdigest(),'bytes':len(raw)}) +print(json.dumps(rows,indent=2)) diff --git a/evidence/indexer-255-token-v3-20261030/validator/evaluate_development.py b/evidence/indexer-255-token-v3-20261030/validator/evaluate_development.py new file mode 100644 index 000000000..4c1f5de54 --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/validator/evaluate_development.py @@ -0,0 +1,32 @@ +import gzip, hashlib, json, sys +from pathlib import Path +root, prep, out, arm = map(Path, sys.argv[1:]) +root=root.resolve() +sys.path.insert(0, str(root/'src')) +from inferrsi.knowledge_operations import KnowledgeReader +from inferrsi._skill_selection import indexed_skill +rows=[] +for filename in ('development.json','negative-controls.json'): + for case in json.loads((prep/'semantic-cases'/filename).read_text())['cases']: + reader=KnowledgeReader() + for path, expected in case.get('source_sha256',{}).items(): + assert hashlib.sha256((root/path).read_bytes()).hexdigest()==expected + options=dict(query=case['query'],relevance=case['relevance'],limit=case['limit'],context_bytes=case['context_bytes'],**case['filters']) + if str(arm)=='token': options['relevance_method']='weighted-token-v3' + result=reader.collect(**options) + ids=[i['id'] for i in result['items']] + rank=next((i+1 for i,x in enumerate(ids) if x in case['relevant_ids']),None) + complete=[] + for context in result['contexts']: + bundle=reader.read(context['ref'],revision=context['revision']) + view={**bundle.view(),'authority':'package'} + _,required,_=indexed_skill(reader.records()[view['id']],view,bundle.content) + emitted={r['path']:r for r in context['resources']} + assert all(p in emitted for p in required) + assert all(r['content'].encode()==bundle.content[p] and r['sha256']==hashlib.sha256(bundle.content[p]).hexdigest() for p,r in emitted.items()) + if view['id'] in case['relevant_ids']: complete.append(view['id']) + raw=json.dumps(result,ensure_ascii=False,sort_keys=True).encode() + (out/(str(arm)+'-'+case['id']+'.json.gz')).write_bytes(gzip.compress(raw,mtime=0)) + rows.append(dict(id=case['id'],rank=rank,ids=ids,complete_relevant=complete,output_bytes=len(raw),context_bytes=result['context_bytes'],context_omissions=result['context_omissions'],sha256=hashlib.sha256(raw).hexdigest())) +(out/(str(arm)+'-summary.json')).write_text(json.dumps(rows,indent=2)+'\n') +print(json.dumps([{'id':r['id'],'rank':r['rank'],'complete':len(r['complete_relevant'])} for r in rows])) diff --git a/evidence/indexer-255-token-v3-20261030/validator/inputs/semantic-cases/development.json b/evidence/indexer-255-token-v3-20261030/validator/inputs/semantic-cases/development.json new file mode 100644 index 000000000..d99b1ad5d --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/validator/inputs/semantic-cases/development.json @@ -0,0 +1,219 @@ +{ + "schema_version": 1, + "corpus_sha": "1147c5bb41724826b64ad13cebc8a4dce5cc278b", + "label_scope": "Independently source-reviewed partial positives. No retrieval calls, scoring implementation, observed rankings or benchmark outputs used to select queries or labels.", + "status": "DRAFT_PENDING_INDEPENDENT_REVIEW", + "contract": { + "positive_metric": "Hit@5 and MRR@5 over listed known-positive IDs; not complete precision/recall.", + "negative_metric": "Nongating out-of-domain diagnostic: report returned IDs/count separately; semantic absence does not imply lexical zero matches. Operational no-match and incompatible-filter gates are separate unchanged-contract checks, not these cases.", + "delivery": "Listed required_resource_suffixes are minimum positive anchors, not the complete required set. Audit all resources in the pinned indexed_skill required set, including references/licenses, plus matching revision. Distinguish retrieval miss, budget omission, and incomplete-resource delivery.", + "heldout": "Do not reveal heldout wording, labels or family details to implementer until candidate runtime/code hash freeze.", + "limitations": "Twelve deliberately selected diagnostic positives, two easy out-of-domain controls; no representativeness, production generalization or agent efficacy claim. Family disjointness is a manually chosen intent taxonomy, not universal independence." + }, + "cases": [ + { + "id": "freshdev01", + "split": "development", + "intent_family": "split_backward_optimizer_validation", + "query": "", + "relevance": "Our pipeline can delay parameter-gradient work after input gradients. Can we also remove the optimizer synchronization point if clipping sometimes fails, or must we keep a rollback path?", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.common.skills.assess_split_backward_schedule" + ], + "forbidden_ids": [], + "required_resource_suffixes": [ + "methods/assess-split-backward-schedule/SKILL.md" + ], + "evidence_paths": [ + "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md" + ], + "source_sha256": { + "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md": "1e64f28ea71672fb53c56f6108b87864738c6d03d6dcd32eedaef9e0625e96e3" + }, + "source_passages": [ + { + "path": "src/inferrsi/data/skill_library/methods/assess-split-backward-schedule/SKILL.md", + "start_line": 20, + "end_line": 20, + "quote": "Removing a barrier must preserve global norm/clipping/NaN decisions, rollback/re-update, and correctness of work dependent on a tentative update." + } + ], + "rationale": "The intent is optimizer validation legality under split backward, not persistent checkpoint recovery or asynchronous snapshot reuse.", + "limitations": "Partial positive label only: other skills are unjudged, not irrelevant. Source-derived candidate guidance does not establish runtime validity or downstream task success.", + "required_resource_suffixes_scope": "minimum anchors only; complete indexed_skill required-resource set must also be audited" + }, + { + "id": "freshdev02", + "split": "development", + "intent_family": "heterogeneous_training_mesh_partition", + "query": "", + "relevance": "Some training stages are activation-heavy and others weight-heavy. How should I compare unequal device groups and stage boundaries without ignoring transfers between different shard layouts?", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.common.skills.choose_parallel_training_plan" + ], + "forbidden_ids": [], + "required_resource_suffixes": [ + "methods/choose-parallel-training-plan/SKILL.md" + ], + "evidence_paths": [ + "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md" + ], + "source_sha256": { + "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md": "25d04aed9782b2a8ff1186828becdc7fce93c4418039ceaef9b8dcc5ec9bcd24" + }, + "source_passages": [ + { + "path": "src/inferrsi/data/skill_library/methods/choose-parallel-training-plan/SKILL.md", + "start_line": 16, + "end_line": 16, + "quote": "Consider uneven stage boundaries and mesh sizes where activation/weight intensity changes." + } + ], + "rationale": "The decision is stage/mesh and sharding placement, not B/W execution order or CPU quota.", + "limitations": "Partial positive label only: other skills are unjudged, not irrelevant. Source-derived candidate guidance does not establish runtime validity or downstream task success.", + "required_resource_suffixes_scope": "minimum anchors only; complete indexed_skill required-resource set must also be audited" + }, + { + "id": "freshdev03", + "split": "development", + "intent_family": "remote_training_sample_lookahead", + "query": "", + "relevance": "Increasing Mosaic sample look-ahead made workers download the same shards repeatedly. How do I decide whether more prefetch helps when the local shard cache is tight?", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.common.skills.training_input_pipeline" + ], + "forbidden_ids": [], + "required_resource_suffixes": [ + "methods/training-input-pipeline/SKILL.md" + ], + "evidence_paths": [ + "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md" + ], + "source_sha256": { + "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md": "64bd2c6b795d966045d1d3cca55976d3d49b08b06df7c7e06a70c2824cf55b7e" + }, + "source_passages": [ + { + "path": "src/inferrsi/data/skill_library/methods/training-input-pipeline/SKILL.md", + "start_line": 61, + "end_line": 62, + "quote": "More look-ahead can therefore increase churn when the local\ncache cannot retain the working set." + } + ], + "rationale": "The query concerns producer sample lookahead and shard eviction, not measuring disk utilization, page-cache effects or checkpoint loading.", + "limitations": "Partial positive label only: other skills are unjudged, not irrelevant. Source-derived candidate guidance does not establish runtime validity or downstream task success.", + "required_resource_suffixes_scope": "minimum anchors only; complete indexed_skill required-resource set must also be audited" + }, + { + "id": "freshdev04", + "split": "development", + "intent_family": "conflicting_skill_trace_lessons", + "query": "", + "relevance": "Two execution traces recommend opposite changes to the same agent procedure. How can I preserve both lessons instead of letting the more frequent outcome overwrite the other?", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.common.skills.skill_evolution_playbook" + ], + "forbidden_ids": [], + "required_resource_suffixes": [ + "methods/skill-evolution-playbook/SKILL.md" + ], + "evidence_paths": [ + "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md" + ], + "source_sha256": { + "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md": "9cd37f21fe0d862069e4798edca8a04d59f8b840e55d521901594a63dddd9ba3" + }, + "source_passages": [ + { + "path": "src/inferrsi/data/skill_library/methods/skill-evolution-playbook/SKILL.md", + "start_line": 56, + "end_line": 57, + "quote": "Do not resolve\n by voting on whichever outcome appeared more often in the pool." + } + ], + "rationale": "Direct conditional lesson reconciliation and inconclusive escalation; not current candidate acceptance or retrieval rankings.", + "limitations": "Partial positive label only: other skills are unjudged, not irrelevant. Source-derived candidate guidance does not establish runtime validity or downstream task success.", + "required_resource_suffixes_scope": "minimum anchors only; complete indexed_skill required-resource set must also be audited" + }, + { + "id": "freshdev05", + "split": "development", + "intent_family": "hopper_instruction_applicability", + "query": "", + "relevance": "A kernel recipe uses tcgen05 and tensor memory, but my deployment is H100. Does an SM90 label make that recipe usable, and what compiled instruction path must I verify?", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.cuda.skills.reason_hopper_kernels" + ], + "forbidden_ids": [], + "required_resource_suffixes": [ + "hardware/cuda/reason-hopper-kernels/SKILL.md" + ], + "evidence_paths": [ + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md" + ], + "source_sha256": { + "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md": "cec4e3aa1eb64c31c54b5f5ca40af2b61b7919d94be4174968581d5f752c0538" + }, + "source_passages": [ + { + "path": "src/inferrsi/data/skill_library/hardware/cuda/reason-hopper-kernels/SKILL.md", + "start_line": 30, + "end_line": 32, + "quote": "Do not map\nBlackwell tcgen05, TMEM, hardware block-scaled MMA, CLC or two-SM MMA recipes to\nHopper." + } + ], + "rationale": "Architecture-specific instruction applicability, not generic kernel tiling correctness.", + "limitations": "Partial positive label only: other skills are unjudged, not irrelevant. Source-derived candidate guidance does not establish runtime validity or downstream task success.", + "required_resource_suffixes_scope": "minimum anchors only; complete indexed_skill required-resource set must also be audited" + }, + { + "id": "freshdev06", + "split": "development", + "intent_family": "simulation_near_capacity_calibration", + "query": "", + "relevance": "A serving simulator matches operator times at light load. Is that sufficient to choose a deployment that meets tail latency as arrivals approach saturation?", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [ + "inferrsi.common.skills.calibrate_serving_simulation" + ], + "forbidden_ids": [], + "required_resource_suffixes": [ + "methods/calibrate-serving-simulation/SKILL.md" + ], + "evidence_paths": [ + "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md" + ], + "source_sha256": { + "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md": "ac773c6f1c110a664dfca28a3ae9129b10c07614e60ca3640dea776220e7b352" + }, + "source_passages": [ + { + "path": "src/inferrsi/data/skill_library/methods/calibrate-serving-simulation/SKILL.md", + "start_line": 19, + "end_line": 19, + "quote": "Include the intended load regime and capacity boundary; do not substitute a low-load check for an SLO claim." + } + ], + "rationale": "Empirical predictor calibration domain and queueing risk, not fixed-workload timing-denominator audit.", + "limitations": "Partial positive label only: other skills are unjudged, not irrelevant. Source-derived candidate guidance does not establish runtime validity or downstream task success.", + "required_resource_suffixes_scope": "minimum anchors only; complete indexed_skill required-resource set must also be audited" + } + ] +} diff --git a/evidence/indexer-255-token-v3-20261030/validator/inputs/semantic-cases/negative-controls.json b/evidence/indexer-255-token-v3-20261030/validator/inputs/semantic-cases/negative-controls.json new file mode 100644 index 000000000..bee802cee --- /dev/null +++ b/evidence/indexer-255-token-v3-20261030/validator/inputs/semantic-cases/negative-controls.json @@ -0,0 +1,49 @@ +{ + "schema_version": 1, + "corpus_sha": "1147c5bb41724826b64ad13cebc8a4dce5cc278b", + "label_scope": "Independently source-reviewed partial positives. No retrieval calls, scoring implementation, observed rankings or benchmark outputs used to select queries or labels.", + "status": "DRAFT_PENDING_INDEPENDENT_REVIEW", + "contract": { + "positive_metric": "Hit@5 and MRR@5 over listed known-positive IDs; not complete precision/recall.", + "negative_metric": "Nongating out-of-domain diagnostic: report returned IDs/count separately; semantic absence does not imply lexical zero matches. Operational no-match and incompatible-filter gates are separate unchanged-contract checks, not these cases.", + "delivery": "Listed required_resource_suffixes are minimum positive anchors, not the complete required set. Audit all resources in the pinned indexed_skill required set, including references/licenses, plus matching revision. Distinguish retrieval miss, budget omission, and incomplete-resource delivery.", + "heldout": "Do not reveal heldout wording, labels or family details to implementer until candidate runtime/code hash freeze.", + "limitations": "Twelve deliberately selected diagnostic positives, two easy out-of-domain controls; no representativeness, production generalization or agent efficacy claim. Family disjointness is a manually chosen intent taxonomy, not universal independence." + }, + "cases": [ + { + "id": "freshnegative01", + "split": "negative_control", + "intent_family": "dental_care", + "query": "", + "relevance": "What toothpaste should I use for sensitive teeth after a dental cleaning?", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [], + "forbidden_ids": [], + "no_known_relevant_rationale": "The reviewed skill owners concern ML systems, hardware, repository research and evidence workflows; none advertises dentistry, toothpaste selection or patient aftercare. Medical advice is outside the known skill corpus.", + "evidence_inventory": "reviewed-corpus-inventory.json", + "expected_diagnostic": "No relevant known skill; record any returned IDs without claiming exhaustive irrelevance of every possible corpus passage.", + "limitations": "Easy out-of-domain negative, not a near-domain hard negative or calibrated abstention test.", + "gating": false + }, + { + "id": "freshnegative02", + "split": "negative_control", + "intent_family": "municipal_parking_permit", + "query": "", + "relevance": "How do I renew a residential street parking permit after moving to a new address?", + "filters": {}, + "limit": 5, + "context_bytes": 262144, + "relevant_ids": [], + "forbidden_ids": [], + "no_known_relevant_rationale": "No reviewed skill owner advertises municipal permits, residence eligibility, address changes or parking administration. Repository configuration and hardware admission are not relevant substitutes.", + "evidence_inventory": "reviewed-corpus-inventory.json", + "expected_diagnostic": "No relevant known skill; record any returned IDs without claiming exhaustive irrelevance of every possible corpus passage.", + "limitations": "Easy out-of-domain negative, not a near-domain hard negative or calibrated abstention test.", + "gating": false + } + ] +} diff --git a/tests/test_evidence_inventory.py b/tests/test_evidence_inventory.py index 754ddac6d..dc5fd64f6 100644 --- a/tests/test_evidence_inventory.py +++ b/tests/test_evidence_inventory.py @@ -145,6 +145,6 @@ def test_checked_in_inventory(): repo = Path(__file__).resolve().parents[1] result = validate_inventory(repo) assert result["errors"] == [] - assert len(result["bundles"]) == 11 - assert result["checked_hashes"] == 232 + assert len(result["bundles"]) == 14 + assert result["checked_hashes"] == 311 assert result["unavailable"] # Historical raw artifacts are not in this checkout. diff --git a/tools/evidence_inventory.json b/tools/evidence_inventory.json index b4b86ecef..c0fb3f82e 100644 --- a/tools/evidence_inventory.json +++ b/tools/evidence_inventory.json @@ -1,6 +1,18 @@ { "version": 1, "bundles": { + "indexer-255-token-v3-20261030": { + "kind": "campaign", + "manifests": [{"path": "plan/publication-integrity.json", "schema": "bundle-sha256"}] + }, + "indexer-255-blind-cue-20261030": { + "kind": "campaign", + "manifests": [{"path": "plan/publication-integrity.json", "schema": "bundle-sha256"}] + }, + "indexer-255-covered-cue-20261030": { + "kind": "campaign", + "manifests": [{"path": "plan/publication-integrity.json", "schema": "bundle-sha256"}] + }, "indexer-255-token-v2-20261030": { "kind": "campaign", "manifests": [{"path": "plan/integrity.json", "schema": "bundle-sha256"}]