Repository navigation
fix: preserve grounded specs across video shards - #2638
Draft
groupthinking wants to merge 5 commits into
Draft
groupthinking wants to merge 5 commits into
groupthinking wants to merge 5 commits into
Conversation
Contributor
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Contributor
|
Important Draft PR not reviewedDraft PRs are not automatically reviewed by default.
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. |
github-actions
Bot
added
javascript
Contributor
|
Please add a See governance: #898 |
Contributor
✅ E2E Test Results: ALL TESTS PASSED
Test Output |
![vercel[bot]](https://avatars.githubusercontent.com/in/8329?s=60&v=4){kind=small}
Copilot started work to fix failing Actions workflows on behalf of
groupthinking
November 10, 2026 08:55
Contributor
There was a problem hiding this comment.
🟡 Changes recommended
Unresolved merge defects can silently drop or mislink requirements and invalidate entire grounded specs through ID collisions, schema-limit overflow, and incorrect citation rebasing.
7 open issues
What changed in this PR
Extends UVAI’s Video Pack extraction to carry grounded specifications through multi-shard video processing.
Changes:
- Requests partial grounded specs with section-local citation indices from shard workers.
- Combines available candidates and rebases citations against merged transcript and visual evidence.
- Adds fixture-based regression coverage and updates sectional prompt assertions.
| File | Description |
|---|---|
apps/web/src/test/real-aujzb1d-fag-evidence.ts |
Adds captured evidence and a grounded-spec fixture. |
apps/web/src/lib/video-pack-extractor.ts |
Requests partial grounded specs from sectional workers. |
apps/web/src/lib/video-pack-extract-merge.ts |
Merges grounded candidates and rebases citation indices. |
apps/web/src/lib/__tests__/video-pack-multishard-grounded-spec.test.ts |
Exercises multi-shard extraction through final pack binding. |
apps/web/src/lib/__tests__/video-pack-extractor.test.ts |
Updates expectations for the sectional extraction prompt. |
Review effort: Max
Comment on lines
+184
to
+185
| requirements: dedupeBy(rebased.flatMap((candidate) => candidate.requirements), (requirement) => requirement.id || requirement.title), | ||
| acceptanceCriteria: dedupeBy(rebased.flatMap((candidate) => candidate.acceptanceCriteria), (criterion) => criterion.id), |
Contributor
✅ E2E Test Results: ALL TESTS PASSED
Test Output |
Contributor
|
Deployment failed for project v0-uvai with the following error: Learn More: https://vercel.com/docs/concepts/projects/project-configuration |
Contributor
🔍 PR Validation |
This branch had an error being deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #2639
Multi-shard extraction now retains available grounded specifications and their exact source references, including independent shards that reuse local IDs. This follow-up completes the shared sectional prompt constraints and chooses complete application metadata from a later candidate when the first candidate has empty fields.
Current correction
Head:
2f9649262539638e75104c6c318a69da72dfd326. Tree:a5227c5811fc773c9dfba6b0d180dcf533a8e51b, identical between the local staged merge and GitHub's created tree. This history-preserving integration retains prior head5d4375e2ef48261b44877a8ab3ea29189ac99373and current main512a2558f73d6281b6a880548bedf140b7e30292as parents. Application source and regression changes are unchanged by this integration; current main's manifests, lockfile and security workflow are retained byte-for-byte.aggregate-limitdiagnostic with the collection and count. No source-grounded rows are silently truncated.codex/fix-multishard-grounded-specthroughgit.deploymentEnabledin root and app Vercel configuration. The first follow-up used an invalidgithub.enabledobject; Vercel rejected it before building withgithub.enabled should be boolean. Configuration-only commit5d4375e2ef48261b44877a8ab3ea29189ac99373corrects that shape to the documented branch map: https://vercel.com/docs/project-configuration/git-configuration. This fixes the configuration error without enabling a preview.Earlier head
7dcc3642900aed3a700c898220d8010d4dd781ddalready added shard-local ID namespaces and reference rewriting, exact-case evidence preservation, raw-to-parsed citation rebasing, and shard-time-bound validation. Their regressions passed again with this correction.Verification — November 12, 2026 (America/Chicago)
The previous head's Security Scan 43346308028 failed the real high-severity npm-audit gate: 25 high, 22 moderate, zero critical. Its frozen install passed. That failure was caused by the old dependency baseline, including affected Undici, Zod, body-parser, Vite and other packages; it was not the expected missing-preview failure.
The clean integration imports current main's already-reviewed dependency fixes and preserves exactly its full-graph
npm audit --audit-level=highpolicy. No exception, severity downgrade, dev-dependency exclusion, or CI weakening is added. Relative to main, the PR remains eight files with the same 517 additions/30 deletions: the existing source/tests/evidence fixture and two branch deployment guards.Executed against the exact integrated tree:
npx --yes npm@10.8.0 ci --ignore-scripts --legacy-peer-deps --no-audit --no-fundcompleted, exit 0. The reviewed existing Workflow postinstall patch was run explicitly.npm --workspace=apps/web test -- src/lib/__tests__/video-pack-multishard-grounded-spec.test.ts src/lib/__tests__/video-pack-extractor.test.ts src/lib/__tests__/video-pack-extract-merge.test.ts src/lib/__tests__/grounded-spec-pipeline.test.ts src/lib/__tests__/video-pack-consolidator.test.ts src/lib/__tests__/dependency-runtime-compatibility.test.ts src/lib/__tests__/gemini-sdk-continuation.test.ts --maxWorkers=2 --reporter=dot: seven files, 110 tests passed, exit 0.npm --workspace=apps/web run type-check, full web lint, and cached diff whitespace check: exit 0.npx --yes npm@10.8.0 audit --audit-level=high --json: exit 0, zero critical, zero high, 29 moderate. Moderate findings remain; this is not a clean-audit claim.git.deploymentEnabled.a5227c5811fc773c9dfba6b0d180dcf533a8e51bequals GitHub's tree exactly. Existing branch update used an expected-head lease.Security dependency evidence
512a2558f73d6281b6a880548bedf140b7e30292, preserving its direct/transitive classification and floors, including Workflow-owned Undici >=7.30.1, Zod >=4.6.6, body-parser >=2.4.0 and Vite >=8.3.3.Remaining HOLD: aggregate capacity
The review request to keep all 66 requirements usable remains unresolved. The final schema allows at most 64 rows per collection. Three valid 22-row shards therefore produce a blocking diagnostic rather than an available grounded spec. Changing the supported aggregate bound or selecting a bounded partial result requires a separate reviewed decision; this patch preserves the existing bound and fail-closed behavior. The overflow thread remains open.
Seven other review threads are resolved using the implemented corrections and passing regressions. This remains a draft and unmerged, with this branch excluded from automatic deployments in both versioned configurations. No live extraction, paid provider call, server-side action or production acceptance was executed; the regression uses saved source evidence and offline test responses.