Add inert source-backed claims schema foundation#118
Merged
Conversation
Establish storage-only foundation for future source-backed claim packets. Nothing reads or writes this table yet: no extraction, queue, MCP tools, API routes, search fusion, or dashboard — existing search is untouched. - scripts/setup-db-claims*.sql: claims table across all four provider variants (1536/3072/1024/768), dimension-only diffs mirroring the insights schema pattern (inline RLS, generated FTS column, shared update_updated_at trigger). - Claims are chunk-anchored: chunk_id is NOT NULL and offsets are UTF-16 indices into chunks.content, the only basis that round-trips per src/utils/source-span.ts. embedding is a nullable vector(N) column with NO vector index in this PR (provider-safe indexes, incl. the Google 3072d HNSW question, deferred to the retrieval PR). - Light constraints only: status/state/sensitivity CHECK sets, ordered non-negative offsets, confidence in [0,1], no self-supersession. - Add Claim/ClaimStatus/ClaimState/ClaimSensitivity types. - Add claims-schema text-contract smoke test (65 assertions) that keeps the four variants in sync and asserts no vector index is present. - Document the schema in CLAUDE.md feature schema list. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SscVJjuHF4BMezAxW9by5m
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (5)
🚧 Files skipped from review as they are similar to previous changes (5)
WalkthroughFour provider-specific SQL setup scripts are added ( 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@scripts/setup-db-claims.sql`:
- Around line 103-104: The CREATE POLICY statement for claims_deny_anon will
fail on subsequent script runs because the policy already exists. Add a DROP
POLICY IF EXISTS claims_deny_anon ON claims statement immediately before the
CREATE POLICY claims_deny_anon statement to make the policy creation idempotent.
This ensures the policy is safely dropped (if it exists) before being recreated
on each run.
- Around line 35-37: The table is missing a constraint to ensure that the
chunk_id actually belongs to the document_id. Currently, document_id and
chunk_id are validated independently against their respective tables, but there
is no guarantee they form a valid document-chunk pair. Add a composite foreign
key constraint that references both the document_id and chunk_id columns
together against the chunks table to enforce this relationship, ensuring that
only chunk_id values that actually belong to the specified document_id are
allowed in the table.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 9866a068-cc1e-4bfb-a754-6cdecebd4b60
📒 Files selected for processing (7)
CLAUDE.mdscripts/setup-db-claims-google.sqlscripts/setup-db-claims-ollama-v2.sqlscripts/setup-db-claims-ollama.sqlscripts/setup-db-claims.sqlsrc/db/__tests__/claims-schema.test.tssrc/types/database.ts
…/chunk FK - Make RLS policy creation rerunnable: DROP POLICY IF EXISTS before CREATE POLICY claims_deny_anon, so re-applying the script no longer errors (the rest of the script was already idempotent). - Enforce provenance pair integrity at the schema level: a claim's (document_id, chunk_id) must be a real pair, not two independently valid ids. Replace the standalone chunk_id FK with a composite FOREIGN KEY (document_id, chunk_id) REFERENCES chunks(document_id, id) ON DELETE CASCADE. The required unique index on chunks(document_id, id) is created by the claims script itself (chunks.id is already unique, so it always builds), keeping the base setup-db*.sql ingestion path untouched for deployments that don't enable claims. - Extend the schema smoke test to assert the composite FK, its unique index target, and the idempotent policy across all four variants. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SscVJjuHF4BMezAxW9by5m
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
2 participants
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.
Summary
Establishes a storage-only foundation for future source-backed claim packets, the next step after the merged source-span resolver (
src/utils/source-span.ts).This PR is deliberately inert: nothing reads from or writes to the new table yet. No extraction (LLM), queue, MCP tools, API routes, search fusion, or dashboard — existing search behavior is untouched and no new tools are registered.
What's included
scripts/setup-db-claims.sql(1536),-google.sql(3072),-ollama.sql(1024),-ollama-v2.sql(768). Mirrors the insights schema pattern: inline RLS (deny anon/authenticated), generated weighted FTS column (claim > question > quote), and reuse of the sharedupdate_updated_at()trigger.Claim,ClaimStatus,ClaimState,ClaimSensitivityinsrc/types/database.ts.src/db/__tests__/claims-schema.test.ts— a pure text-contract smoke test (65 assertions) modeled onuploads-schema.test.ts, keeping the four variants in sync and asserting that no vector index is present.Key design decisions
chunk_idNOT NULL,ON DELETE CASCADE). The offsets are UTF-16 indices intochunks.content— the only basis that round-trips per the source-span resolver. A claim cannot exist without its anchoring chunk.document_idis kept (NOT NULL, CASCADE) as a denormalized convenience.embedding vector(N)with NO vector index. No embedding pipeline writes claims yet. Provider-safe vector indexes — including the pgvector >2000-dim HNSW question on the Google 3072d variant — are deferred to the future retrieval PR. The schema test enforces the absence of anyhnsw/ivfflatindex.superseded_byself-FK withON DELETE SET NULL). Avoids over-constraining before real extraction exists.Out of scope (future PRs)
LLM/claim extraction, a writer/db module, processing queue, MCP tools, API routes, search/retrieval fusion, visual retrieval, and the project-wide Google 3072d HNSW story.
Verification
pnpm verify:fastpasses locally: lint, typecheck, 416 tests (65 new), security, docs, and tool-sync checks all green. Confirmed inert — no runtimesrc/code references the table.🤖 Generated with Claude Code