Purpose
Make sherlock-search the hybrid query membrane over Prophet Understand / Repo Intelligence artifacts.
Parent epic: SocioProphet/prophet-platform#407
Required capability
Sherlock should query repo-intelligence artifacts using lexical, fuzzy, vector-ready, graph-aware, and evidence-aware ranking. The v0 lane should work with local fixtures before live estate-scale ingestion.
Initial implementation path
- Add
docs/PROPHET_UNDERSTAND_SEARCH.md.
- Define the query contract for repo graph artifacts: nodes, edges, summaries, source anchors, provenance receipts, policy status, and validation results.
- Add fixture queries:
what owns this file?
what depends on this contract?
what tests cover this node?
what changed in this PR impact set?
what policy gates touch this service?
- Add ranking notes for lexical score, semantic/vector placeholder score, graph proximity, evidence quality, freshness, and policy risk.
- Return answer objects with citations/source anchors, confidence, and graph path explanation.
Acceptance criteria
- Search over a fixture artifact returns explainable results with node IDs, paths, graph relationships, and provenance metadata.
- Ranking does not claim semantic certainty where no embedding/provenance exists.
- Query results preserve policy and validation status.
- The search contract can later federate across Lampstand and Sociosphere without schema rewrite.
Non-goals
- Do not reduce Sherlock to Fuse.js-only local search.
- Do not vendor Understand-Anything.
- Do not introduce opaque agent-generated answers without evidence paths.
Purpose
Make
sherlock-searchthe hybrid query membrane over Prophet Understand / Repo Intelligence artifacts.Parent epic: SocioProphet/prophet-platform#407
Required capability
Sherlock should query repo-intelligence artifacts using lexical, fuzzy, vector-ready, graph-aware, and evidence-aware ranking. The v0 lane should work with local fixtures before live estate-scale ingestion.
Initial implementation path
docs/PROPHET_UNDERSTAND_SEARCH.md.what owns this file?what depends on this contract?what tests cover this node?what changed in this PR impact set?what policy gates touch this service?Acceptance criteria
Non-goals