[waleedlatif1]

Summary

• Live search document IDs were live: + base64 of a JSON object naming the reader, scope, account and document (~270-375 chars, ~175-245 tokens). Clients re-read them every turn and copy them verbatim into reads
• New searches mint live2: IDs: provider~account~id~container~kind~thread~revision~binding~checksum, ~65-135 chars (~47-64 tokens)
  • Reader and scope come from the request: the old fields were never an authorization boundary (unsigned), reads still re-resolve the account among the reader's own accounts in scope and re-verify the document
  • A 6-char checksum makes a truncated or edited copy ask for a new search instead of naming another document
  • Only the separator, % and control characters are escaped, so an ID is never longer than the old form and any string encodes
• Old live: IDs still read, with their reader/scope check unchanged
• Every ID comparison (filters, read check, continuation fingerprint, MCP batch dedupe, citation IDs) goes through a canonical document key, so old and new forms of one document match

Type of Change

• Improvement

Testing

• New reference.test.ts + updated application/filters/MCP server tests; 6 new tests seen failing on the previous encoding (truncation, edit, unpaired surrogate, length, mixed-form batch dedupe) and passing now
• Targeted vitest: lib/sim-search/live, lib/knowledge/search, lib/knowledge/mcp, mothership knowledge tools, chat citations, message-content citations (1,356 passed); biome on changed files; type-check, audits and integration in CI
• Paired Claude-as-MCP-client benchmark (Sonnet 5.5, prompt caching, 83 cases × 2, judged; vs two control runs of staging): input tokens per question −15k to −19k, model time −1.0 to −1.5 s, tool and schema errors to zero, correctness and completeness flat or up on both dev and held-out splits

Checklist

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

🤖 Generated with Claude Code

[vercel]

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

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Nov 8, 2026 12:10am UTC

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no new actionable issues were found.

Summary

This PR replaces long live document references with shorter live2: IDs.

• Live search returns shorter document IDs that still work for reads.
• Older and newer live IDs match as the same document.

No new actionable issues were found.

_{Reviews (3) · Last reviewed commit: "fix(search): bind live IDs to their read..." · Reviewed by Greptile}

[cubic-dev-ai]

All reported issues were addressed across 10 files

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

Turn on auto-fix | 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.

_{Turn on auto-fix | 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]

No issues found across 9 files

Confidence score: 5/5

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

_{Turn on auto-fix | Re-trigger cubic}