[AdaWorldAPI]

Purpose

Choose a low-risk cold ONNX importer by comparing pinned Burn onnx-ir and Tract InferenceModel→TypedModel on the same real model. Retain ONNX source provenance and typed method semantics; then lower one independently verified native subgraph through existing ndarray / lance-graph-onnx / Quack. No new ONNX executor.

This is S0 documentation only, not a parser implementation, numerical parity claim or approval for S1–S5.

Files

• Capstone plan
• Claude Code S1/S2 handoff
• Concise board entry

Existing boundaries and related work

• D-NFC #1753, native ONNX leaf #1754/#1758, historical intake #1744/#1750.
• EXO-152 D-NAM owns native method and numerical receipt; EXO-50 receives coordination.
• DT64 #1794 catalogue ≠ ONNX node identity; #1795 source/snapshot/ACL; #1796 directed DR ≠ MatMul; #1797 hypothesis/temporal ≠ model result; #1798 native numerical method leaf; #1799 canonical D-DT64-Q0 source-attested borrowed binding. No duplicate QB/FrameStore.
• #1792 merged: compiled recipe reuse is not stale escaped-Keep reuse.

Review questions / future stages

S1: compare source fidelity, dependency/compile cost, symbolics, external initializers and op versions with optimization off/on.
S2: measure admitted native node-weighted coverage on one fixed BERT/Jina ONNX model; typed refusals.
S3: one complete static subgraph via existing kernels, checked against ORT; never use importer runtime as an invisible fallback.
S4/S5: block-specific numerical certification and full-model forward parity only under separate approval.

Scope and verification

• Docs-only, 3 new Markdown files.
• Related DT64 docs(DT64-1): 64 typed predicate mappings for IAM Odoo SAP #1794–docs(DT64-0): versioned DTO predicate adapter capstone, staged PR board and EXO-50 gates #1799 and canonical Q0 reviewed.
• Model import, ORT comparison and tests: NOT RUN (not within S0).
• Third-party parser/version/dependency selection: OPEN.
• Implementation authorization, runtime binding and full-model support: NOT GRANTED.

Do not merge autonomously. GitHub is the source of truth for this plan; EXO-50 is append-only coordination.

Updated upstream baseline (merged 2026-10-26)

• #1800 merged: TFLM Softmax source-oracle replay/fixtures and certified preparation gate, lab only; numeric precedent, not ONNX model execution.
• #1801 merged: Codec8 code-space/receipt admission, lab only; reused in refusal design, not a production TFLite runtime or binding store.
• Plan, CC handoff and board entry updated on this branch; no reimplementation of those merged tests is proposed.

[coderabbitai]

Important

• 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: c80bcdfc-7c8f-453a-a7b8-3979289bdb7c

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Warning

Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-26T20:59:23.401423Z	1f515d0	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.