Search Rrf
Execute RRF hybrid search combining vector similarity and keyword matching.
Reciprocal Rank Fusion (RRF) combines rankings from both search methods:
- Vector search: Semantic similarity based on embeddings
- Keyword search: Full-text search with language-specific stemming
Results appearing high in both searches are boosted.
Args: request: RRF search request with query and parameters service: Hybrid search service (injected)
Returns: Search results with RRF scores and component breakdowns
Raises: HTTPException: 503 if service not configured, 400 for invalid request
Request Body required
Section titled “Request Body required ”Request model for RRF hybrid search.
object
Search query text
Embedding model name (must match indexed model)
Full-text search language
Weight for vector results
Weight for keyword results
RRF smoothing constant
Max candidates per search type
Maximum results to return
Number of results to skip
Responses
Section titled “ Responses ”Successful Response
Response model for RRF hybrid search.
object
Single result from RRF hybrid search.
object
object
Validation Error