fix(knowledge): bound source titles and tag values so an oversized index row cannot fail a connector sync

[waleedlatif1]

Summary

• a document's filename and text tag columns carry btree indexes, and Postgres refuses an index row past 2704 bytes (SQLSTATE 54000). Reproduced on the real schema: an incompressible 2,700-character filename fails with index row size 2712 exceeds btree version 4 maximum 2704 for index "doc_filename_idx", a 3,200-character tag fails on doc_kb_tag1_lower_idx, 2,600 characters insert. Connector syncs wrote source titles and mapped metadata straight into those columns, and the failure-recording path re-inserts the same title, so one such document escalates to a connector-level failure on every sync and ten of those disable the connector
• one shared bound, MAX_DOCUMENT_INDEXED_TEXT_LENGTH (512 characters, under the ceiling even at four bytes per character) in lib/knowledge/constants.ts
• connector ingress truncates to it (source titles and mapped tag values are machine input with no caller to refuse), cutting by code point so a value never ends in half a surrogate pair
• the document APIs reject over-long input instead: .max() on filename (create, upsert, update) and on tag1..tag7 (create, update), and the upload-metadata tag bound (previously 1000 characters, which still overflows at four bytes per character) now uses the same constant
• not in this PR: doc_filename_idx itself looks close to unused in production (338 scans against millions on the tag indexes); dropping it is a separate, evidence-gated migration

Type of Change

• Bug fix

Testing

• sync-persistence.test.ts: bounded filename on the placeholder path, bounded and intact mapped tag values, code-point cut on a 600-emoji title
• documents.test.ts: at-limit accepted; over-limit filename rejected on create/upsert/update and over-limit tag rejected on create/update, with the descriptive messages
• upload-metadata.test.ts: over-limit tag rejected
• all five new tests fail with the bound neutralized (mutation-checked); 60 tests pass across the three files
• bun run type-check and bun run check:api-validation pass; the mechanism was reproduced on a local database with the real indexes inside rolled-back transactions

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 22, 2026 3:35am UTC

[cubic-dev-ai]

No issues found across 2 files

Confidence score: 5/5

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

_{Re-trigger cubic}

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge, with the previous findings resolved and no new actionable defects identified.

Summary

This PR prevents oversized document filenames and tag values from repeatedly failing knowledge connector synchronization.

• Introduces a shared 512-code-unit bound for indexed document text.
• Truncates connector-supplied titles and mapped tags without splitting surrogate pairs.
• Rejects oversized API and document-service inputs with descriptive validation errors.
• Adds boundary, exact-result, and Unicode coverage for connector and API paths.
• Preserves values already within the accepted limit, addressing the remaining previous-review concern.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Connector source metadata] --> B[Bound title and mapped tags]
  B --> C[Document persistence]
  D[Document API input] --> E[Reject values over shared limit]
  E --> C
  C --> F[PostgreSQL filename and tag indexes]

Loading

_{Reviews (5) · Last reviewed commit: "fix(knowledge): leave values at the limi..."}

[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 2 files

Confidence score: 5/5

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

_{Re-trigger cubic}

[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]

All reported issues were addressed across 7 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]

All reported issues were addressed across 7 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 11 files

Confidence score: 5/5

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

_{Re-trigger cubic}