Jordan-Hall / Jordan-Hall/browser
[P1][DATA-02] Hybrid retrieval with source-scoped authorization
- Dominant language
- No language data
- Stars
- 0
- Forks
- 0
- PR merge metrics
- No merged PRs in 30d
Description
Programme: #1
Epic: #17
## Objective
Build retrieval that combines authoritative relational state, exact full-text search and optional semantic retrieval without leaking private or out-of-scope data into model context.
## Scope
- Relational queries for authoritative entities/permissions/source metadata.
- Full-text index for exact names, IDs, code, quotations and documents.
- Optional embedding/vector index and reranking for semantic retrieval.
- Task-scoped context builder with access/destination checks before retrieval material is emitted.
- Evidence-preserving summaries and chunk references.
- Account/workspace/profile-aware caches and index partitions.
- Deletion/retention hooks for all searchable derivatives.
## Security rules
- Authorization occurs before snippets/embeddings/summaries are exposed to a worker/model.
- Embeddings and cached summaries inherit privacy restrictions.
- Work/personal/private-session indexes remain isolated.
## Acceptance criteria
- [ ] A denied/private record cannot appear through snippets, embedding neighbors, reranker input or cached summaries.
- [ ] Exact-ID/full-text queries prefer authoritative matches over approximate semantic matches.
- [ ] Context bundles preserve evidence links and unresolved constraints.
- [ ] Index updates and deletions are observable and recoverable after crashes.
- [ ] Cloud-context construction applies explicit destination policy before serialization.
- [ ] Retrieval evaluation reports recall, precision, latency and privacy failures separately.
## Dependencies
- DATA-01
- SEC-03
**First phase:** P1
**Maturity target:** P3
**Owner:** workspace-ui-data
## Task issues
- [ ] #351 `DATA-02.T01` — Implement authoritative exact retrieval
- [ ] #352 `DATA-02.T02` — Build incremental full-text indexing
- [ ] #353 `DATA-02.T03` — Add embedding and semantic candidate retrieval
- [ ] #354 `DATA-02.T04` — Implement fusion and scoped reranking
- [ ] #355 `DATA-02.T05` — Build evidence-preserving context bundles
- [ ] #356 `DATA-02.T06` — Implement deletion, cache invalidation and rebuild
- [ ] #357 `DATA-02.T07` — Measure quality and privacy independently
Contributor guide
No contributing guide indexed for this repository
Research direction
No implementation files, tests, or entry points are named. Start by reviewing dependencies DATA-01 and SEC-03, then use task issues #351–#357 to split the work into authoritative retrieval, indexing, semantic retrieval, context bundles, deletion, and evaluation. Done requires satisfying the listed privacy, authorization, evidence, recovery, policy, and measurement criteria.
Written by the indexing model from the issue text.
Assessment
- Domain
- backend-api-design, search, security
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100