Repository navigation
[Task] RSI Dashboard: Optimization Workflow Visualization and Trace Capture #314
Description
Current state: PR #319 was squash-merged on 2026-11-05 at 09:15 UTC, from reviewed head 2218beab into main commit e843b9d31cb86e1fd12d00890c89798b0bedd1fe. The parent Issue remains open. Amy's current-head review now records APPROVED / LGTM and PASS for the incremental software changes, including the missing-run and summary-navigation repairs. It explicitly leaves Issue scope and acceptance ownership separate; it is not a scope-acceptance decision. Pre-merge Ready CI at the reviewed head succeeded; this status does not claim a post-merge check result. Existing owner @david6666666 retains the work; this synchronization checked records and revision metadata, without rerunning tests or performing another code/browser review.
-
Missing-run repair: the reviewer confirms that available run objects or valid events are now required before presenting a run as
recorded; missing/corrupt runs and summaries citing absent outcomes are distinguished. The delta adds five regression cases. The author reports 158 targeted tests on Windows and Linux, Ruff, and synthetic bilingual browser checks. These are scoped evidence, not this synchronization's independent reproduction or whole-PR acceptance. -
P3 summary navigation resolved within the reviewed delta: the finding at
10f1916dconcerned a disabled summary link for an available run without an outcome. The review at2218beabconfirms thatoutcome_not_recordednow permits navigation while preserving its label and warning; missing, unavailable and ambiguous references remain disabled. The added regression exercises the actual renderer in a Node stub. -
Scope amendment recorded; provenance and acceptance remain separate: the original scope record covers English, read-only offline HTML and local Markdown review export. The author-posted amendment records requester direction to include bilingual UI, optional loopback live serving and caller-authored story/context capture, preserving the stated privacy and evidence boundaries. The earlier review requested direct requester confirmation or the original instruction because the amendment is posted on the requester's behalf. No later direct confirmation is present in this Issue's comments; Amy's new software approval expressly leaves scope/acceptance separate. The 09:06 same-head follow-up still requested scope provenance; a later formal APPROVED record and actual merge do not by themselves supply that missing source or close parent acceptance. This synchronization does not accept or reject scope.
-
Current-head checks and Node coverage: Ready CI at
2218beabsucceeded. The coverage follow-up reports that both Python 3.10/3.11 full-check jobs changed from 3,695 passed / 13 skipped at10f1916dto 3,696 passed / 13 skipped at2218beab, with exactly one added test, the Node regression. This is aggregate execution evidence, not a per-test-name log; the same report records 36 local exact-head targeted passes with no skips. This synchronization did not rerun those tests. Amy's earlier approval at9301d29aremains dismissed; the new approval at2218beabis a separate current-head record. It used static review and existing hosted checks, without a new local Node or live-browser run. -
#344 knowledge-use/adoption attribution and #345 Dashboard enhancements remain unassigned proposals. task: Dashboard 增强：每轮改动说明、动态 workflow 与耗时、测试进度 #345 now also records per-service change/optimization summaries, a dynamic workflow and stage/node elapsed times alongside the separate accuracy/performance progress bars. Its recorded authorization covers creating and consolidating requirements, not implementation, merge or closure; first scope, event interfaces and ownership remain undecided. Neither proposal is an accepted addition to this checklist. Caller-authored narrative and capture hashes do not prove authenticated origin, actual model consumption or causal optimization benefit.
-
External Ascend repair merged separately: PR #358 was squash-merged on 2026-11-05 at 14:20:26 UTC from reviewed head
aeb2c7f79f30e336bd1c08ce00df5862d247f1e4as main commitf1c90dffe321729f4fe37fd95bac1598bcf2729e. GitHub now records #356 closed as completed at 14:20:33 UTC; that repair disposition does not close this parent Issue. The slice maps hash-bound external records into native views, separates per-tier/two-round measurements and quality/performance verdicts, and distinguishes disconnected input from missing or inapplicable stages. The mixed-campaign failure-count blocker from Amy's earlier re-review is cleared: native failures/rollbacks use native rows, and a separate card counts recorded external performance FAIL. Amy's current-head approval and zengchuang-hw's independent approval both record LGTM/PASS. The latter reports 100 exact-head Dashboard tests and 20 effective-merge tests / 1 deselected, including mixed native-unqualified/external-FAIL behavior; these are reviewer evidence, not reruns by this synchronization. Reviewed-head Ready CI and policy passed; the merge commit's merge lane and push policy also succeeded. External verdicts do not enter generic verified-gain ranking, and caller-declared environment identity does not certify experiment comparability. No experiment was run here or historical narrative invented. Existing contributor @jiangkuaixue123 and parent owner @david6666666 retain their respective work; broader rollout, the original parent checklist and scope-provenance question remain separate.
Decision needed: Resolve scope provenance and acceptance against the recorded amendment and its source, separately from the new software approval. Separate scope, interface and ownership decisions for #344/#345 remain pending.
Next action: The existing owner supplies the requested scope provenance, records the merged delivery's disposition and supplies the remaining acceptance evidence for the real poor-outcome feedback case. The original checklist remains open; performance acceptance stays a separate issue-scoped decision.
Revision and prior scoped dispositions
- Merged PR's reviewed head:
2218beabfea8e89f4c221dc17a40f1d8e9353ada; API-reported PR base:2baebf6ebb815ebf07d21c41e3b3754858c0798e. The missing-run repair check is bound tofd2bbe768f2247168bffbbf6922f533b1aca7548; the latest summary-navigation review is bound to the observed current head. - The non-voting review at
9301d29asupplied the missing-run reproduction and scope question. Its 183 targeted Windows/Python 3.12.1 tests, Ruff and whitespace checks are historical results, not current-head reruns. - Earlier translation/verdict and round-specific capture findings received scoped repair confirmation at 870e752f; live-delta review at 806e71dd found no new actionable defect in its inspected paths. Amy's later review at 9301d29a cleared extreme-chart/tick handling and clarified loopback binding with SSH forwarding. These dispositions remain historical; that approval is now dismissed.
- Demo recordings are synthetic. The author-reported Chinese live-browser check against an in-progress campaign preserves incomplete attempts as inconclusive; it does not establish real optimization efficacy or the poor-outcome feedback acceptance case. Historical context that was never saved cannot be reconstructed by this delivery.
Parent issue: #316 · InferRSI Roadmap
Goal
Provide an InferRSI dashboard that makes an optimization task observable from diagnosis and candidate search through execution, validation, and review. Correlate each iteration's results with traces to help diagnose why optimization outcomes fall short.
Owner: @david6666666. Corresponds to original roadmap item 3.
Scope
- Workflow visibility: Show task / run status, stages, iterations, candidates, validation results, and available performance metrics.
- Trace capture and correlation: Link optimization steps to their events, logs, tool call records, and existing profiling artifacts so users can trace results back to specific steps and evidence.
- Problem diagnosis: Surface failures, rollbacks, steps with no improvement, and missing evidence. Help identify whether a problem originates in diagnosis, knowledge selection, search, execution, or validation.
- Human review entry point: Support inspecting and referencing traces so findings and recommendations can feed back into the framework, connecting to the shared feedback loop in roadmap item 7.
Build the first version around the existing RSI workflow and evidence artifacts. The owner will refine integration points, trace fields, and the interface design in this issue.
Acceptance Criteria
- Show an optimization task's stages and iteration results as a connected workflow.
- Make key steps traceable to the corresponding traces, logs, and reports, with explicit indicators for missing records.
- Demonstrate “inspect the workflow → diagnose the problem → produce framework improvement feedback” using a case with poor optimization outcomes.
- Provide setup, integration, and usage instructions, including the first version's coverage and limitations.
Activity
[-][Task] RSI Dashboard：优化流程可视化与 Trace 捕捉[/-][+][Task] RSI Dashboard: Optimization Workflow Visualization and Trace Capture[/+]agent-ci-infra-bot commentedon Nov 2, 2026
意图：为 InferRSI 做一个 Dashboard，把一次优化任务从诊断、候选搜索、执行、验证到复盘的阶段与每轮结果可视化，并与 Trace/日志/证据关联，用于定位效果不佳的原因和支撑人工复盘反馈。结论：需信息 + 需人类决策——Issue 自己把接入点、Trace 字段与界面方案留给负责人在本 Issue 内细化，且与 #61「不假设大范围 GUI」冲突，当前没有可派发的有界切片。核对了仓库记录模型与现有报告入口，未执行任何测试（这是 Issue 评审，无代码改动）。
Labels: kind:task, kind:rfc, area:workbench, area:usability（路线图下的交付任务，但当前仍是待细化的设计，涉及 workbench 证据/报告与人工复盘可用性）
承载形态与产品方向未定，且与已生效的路线图约束冲突
pyproject.toml:12dependencies = []），报告是单文件静态 HTML（src/inferrsi/harness/reporting.py:11），只读入口是 CLIstatus/report（src/inferrsi/cli.py:275）。服务化会引入新的依赖、端口、进程清理与安全边界，这不是实现者能自行选择的。Trace 字段与采集点未定义；现有记录里不存在 Issue 描述的五个阶段
src/inferrsi/harness/runner.py:43（events.jsonl）与:46/:51（status.json）只记录source_snapshot、measurement、quality_validation、complete、crashed这几类事件（runner.py:122-205的event(...)调用）；src/inferrsi/runspec.py:10-31的 RunSpec 只有baseline/candidate_repo与repetitions，没有「轮次」「候选集」概念；src/inferrsi/model/anthropic_agent.py:412-435、:556-581另有一套agent.request/agent.response/agent.tool/agent.context事件，与 workbench 的 run 事件没有共同的 run/尝试 ID。缺少仓库要求的 accepted scope 与 scoped-task 记录字段
AGENTS.md:117「Work Issue-first: confirm the governing Issue's accepted scope before changing code…」；docs/agent-collaboration.md:15「Work follows accepted scope and explicit or standing delegation. Record or link that authority.」；.github/ISSUE_TEMPLATE/scoped-task.md要求 Current state / Decision needed / Next action 以及 Ownership and authorization（含 Status、Accepted decision）。范围过大且混合多个交付物，建议拆分
docs/agent-collaboration.md:15之后的 Working rules（每个 PR 应是有界交付切片）。验收演示的数据来源与隐私边界未说明
docs/privacy-security.md:10-24（日志、报告、证据都不得含真实主机、用户名、绝对路径与凭据，需脱敏）。report.html目前会为 run 目录下所有文件建链，src/inferrsi/harness/reporting.py:26-30，这对「证据浏览」也是一个现成的泄露面）。evidence/的案例与允许展示的字段，明确脱敏与只读边界，并把「缺失证据」的展示方式写成可检查的验收条件。Reviewed commit: 2baebf6
Coverage: Snapshot: 2637 files, 42149780 bytes; skipped 0 entries; Initial model context: 28 files; other accepted files available through tools; Git workspace contains a fixed snapshot, not full upstream history
david6666666 commentedon Nov 2, 2026
The request owner confirmed this implementation scope for #314:
This confirms the dashboard implementation scope. A performance acceptance claim remains a separate issue-scoped decision with its own pinned baseline, workload, quality gates, and evidence.
Add campaign optimization story with captured knowledge context
david6666666 commentedon Nov 5, 2026
Requester-directed scope amendment for #314
The requester subsequently asked to expand the previously recorded scope for the Dashboard in this same delivery. This agent records that follow-up direction on the requester's behalf:
This follow-up supersedes the earlier English-only and no-service exclusions for this issue; the offline option and other stated privacy and evidence boundaries remain. It records the requester's direction and does not by itself establish independent review, campaign acceptance, or authorization to close the issue. Refs #319.
Add offline and live optimization workflow dashboard (#319)