Searching

Search methods

// Hybrid search (default: semantic + BM25 via RRF)
const results = await indexer.search(query, options?);

// Semantic only
const results = await indexer.searchSemantic(query, options?);

// Lexical / BM25 only
const results = await indexer.searchLexical(query, options?);

// Hybrid + graph-context expansion
const results = await indexer.searchWithContext(query, options?);

SearchOptions

{
  limit?: number,                          // default: 10
  language?: Language | Language[],        // filter by language
  types?: ChunkType[],                     // filter by chunk type
  filePath?: string,                       // path filter (glob patterns supported)
  hybrid?: boolean,                        // default: true when embeddings available
  rerank?: boolean,                        // apply reranking after RRF; default: true when rerankingFunction set
  rrfK?: number,                           // RRF rank constant k; default: 45 (tuned for code retrieval)
}

SearchWithContextOptions

Extends SearchOptions:

{
  graphDepth?: number,                     // BFS hops to expand — default: 1
  graphRelationTypes?: RelationshipType[], // edge types to follow — default: all
}

SearchResult

Results are always returned sorted by relevance (best first). The numeric score used internally for ordering is not exposed — it is not comparable across search paths (semantic / lexical / hybrid / reranked). Use matchType to know which path produced a given result.

{
  chunk: CodeChunk,
  matchType: 'semantic' | 'lexical' | 'hybrid',
}