improvement(knowledge): read the source card with the result

[waleedlatif1]

Summary

• Every search re-read the returned documents after retrieval to fetch the name, URL, modified time and connector type of each source card, under the same predicate hydration had just applied (7–82 ms and one round trip per search). Those fields now ride on the hydrated row: hydration joins the connector, and the application maps the row directly. getDocumentMetadataByIds stays for the v1 route and document reads.
• The workspace keyword hydration still scored results on the original vector, which lives out of line in toast storage; it now scores on the projection like the other legs.

Type of Change

• Improvement

Testing

• Application tests updated to feed the source-card fields on the hydrated rows; the case that dropped a passage between hydration and the metadata re-read is gone with the re-read, since hydration's predicate is the gate. The hydration test asserts the joined fields.
• bun run lint, check:audits, type-check pass; 3,311 tests pass across knowledge and the search routes.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Sep 21, 2026 5:08am UTC

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge; the only post-review change restores the shorter Tin readiness-cache window requested by the prior review.

Summary

This PR removes the second source-card metadata read by carrying document and connector fields through hydrated search rows, scores workspace keyword results using the projection vector, and restores the Tin readiness cache to one minute.

• Joins source-card metadata into vector, tag-only, and keyword hydration queries.
• Maps hydrated source metadata directly into application search results.
• Uses the projected embedding for keyword-result distance calculations.
• Updates diagnostics and tests for the single-read retrieval flow.
• Restores the shorter Tin readiness-cache window after the previous review.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Search candidates] --> B[Hydrate visible documents]
  B --> C[Join document metadata]
  C --> D[Join connector type]
  D --> E[Optional reranking]
  E --> F[Map hydrated rows to results]

Loading

_{Reviews (2) · Last reviewed commit: "improvement(knowledge): keep the Tin rea..."}

[cubic-dev-ai]

All reported issues were addressed across 6 files

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

No issues found across 5 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Re-trigger cubic}