Dedupe-style search across the project's
FEATURE_REQUEST tickets. Uses OpenSearch fuzzy match + Voyage AI
rerank — returns the most semantically similar matches to the
query argument, not literal keyword hits. Use this before filing
a new feature request to avoid duplicates.
Returns { matches: [...] } with at most 5 reranked matches, best first, each with its rerank score.
Archived feature requests are excluded, and so are hidden lanes unless includeHiddenLanes=true: those
matches come back with public: false and no status label or share link (new requests land in the
hidden intake lane, so an agent deduping before it files one needs them). When the reranker is
unavailable the matches are unranked (score: null) and ranked is false.
Bearer <token>.minScore=0 is passed.0 disables reranking and returns up to limit candidates.public: false (default: false).Ok