Skip to content

fix(noema): restore null synchronous model timeout - #1893

Closed
seonghobae wants to merge 3 commits into
mainfrom
fix/noema-null-model-timeout-contract
Closed

fix(noema): restore null synchronous model timeout#1893
seonghobae wants to merge 3 commits into
mainfrom
fix/noema-null-model-timeout-contract

Conversation

@seonghobae

@seonghobae seonghobae commented Sep 5, 2026

Copy link
Copy Markdown
Contributor

Root cause

Protected .github/main@bc59c07c448dcfa1b8fbc64b601550d232697f24 changed Noema's synchronous model execution from the established null-timeout contract to a 15-minute step deadline plus a 30-minute job deadline. The same commit rewrote docs/product-goal-directive.md from accepting long-running OpenCode/Strix/Noema model execution to requiring completion within 900 seconds.

That is not an administrative-only wait bound. Prepare Noema model verdict calls the contextual-orchestrator-backed model synchronously, so a step/job wall-clock cap terminates reasoning/stream/tool-call execution solely because elapsed time expired. Noema must distinguish provider/session termination and explicit administrative cancellation from inference elapsed time; provider/model routing remains owned by contextual-orchestrator.

RED → causal repair

  • RED 3e1c5f651293c3aed3f2676269b6a22e05f134c3 restores the pre-regression executable contract in tests/test_noema_orchestrator_workflow_contract.py while leaving the 900-second production timeout in place.
  • production f51960058d12a5c44d799ccb76ce2f6fe8b98c6b restores the pre-fix(noema): bound model runtime to 900 seconds #1890 Noema workflow/docs/test surfaces: no noema-review job deadline, no Prepare Noema model verdict step deadline, and the long-running/null-timeout directive.
  • protected main then advanced through 58c7b045af9c095106e76a233037d19deb2db3c7 (fix(opencode): bound model runtime to 900 seconds #1892) on unrelated OpenCode/Strix contract files. Ordinary two-parent merge 1bd22d048c9aff38ab05cc5f55633e22f2e002a7 non-force restacks the Noema repair on that exact protected head while retaining fix(opencode): bound model runtime to 900 seconds #1892's tree. Fresh compare is ahead with merge-base exactly 58c7b045af9c095106e76a233037d19deb2db3c7; only the four Noema timeout surfaces differ.

The repair intentionally does not change Strix's separately merged #1889 state, provider/model selection, credentials, orchestrator/free, runner selectors, queue policy, security/quarantine/outbound authority, or any product repository domain truth.

Current exact authority

  • protected base: main@58c7b045af9c095106e76a233037d19deb2db3c7;
  • exact PR head: 1bd22d048c9aff38ab05cc5f55633e22f2e002a7;
  • lifecycle: open / Draft / mergeable;
  • exact-head Python Security 33946322090 — queued/non-passing;
  • Security Scan 33946322078 — queued/non-passing;
  • CodeQL PR 33946321978 — pending/non-passing;
  • Agent Review Runtime Quality CI 33946321956 — pending/non-passing;
  • SAST Semgrep 33946322004 — queued/non-passing.

Verification discipline

Keep Draft until exact-head central CI/security/reviewer/coverage/workflow-contract evidence is terminal GREEN with zero valid unresolved findings. If protected main advances, non-force restack and re-audit only the intervening delta. Do not reintroduce a synchronous Noema inference wall-clock bound as a queue-starvation workaround; shared-runner starvation remains the organization control-plane problem tracked separately.

@coderabbitai

coderabbitai Bot commented Sep 5, 2026

Copy link
Copy Markdown

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

  • Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

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

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

@seonghobae seonghobae self-assigned this Sep 5, 2026

Copy link
Copy Markdown
Contributor Author

Protected main@f590a8989ffe133c3caa896bf9c3bd41feca8fc3 merged #1891 with tree c1386cf7e754dc77cc0c70f29a2e510b0cfbd213. This PR's exact head 1bd22d048c9aff38ab05cc5f55633e22f2e002a7 has the identical tree SHA, so every source/test/docs delta from the RED→repair lane is already present in protected truth. Closing this duplicate is therefore verified full succession, not abandonment of an unmerged delta.

@seonghobae seonghobae closed this Sep 5, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant