Postgres Hybrid Search

Uses Reciprocal Rank Fusion (RRF) approach to combine full-text search and vector search.

Assumes your primary key is "id", vector column is "embedding" and tsvector column is "content", but is flexible enough to work with any table or column name

Options:

• rrf_k: The number of top results to consider for RRF (default: 50)
• vector_column: The column name of the vector field (default: :embedding)
• vector_weight: The weight of the vector search (default: 1.0)
• tsvector_column: The column name of the tsvector field (default: :content)
• tsvector_weight: The weight of the tsvector search (default: 1.0)
• select_columns: The columns to select from the schema (default: all columns)
• limit: The number of results to return (default: 100)

Keyword Search

Uses Postgres full-text search (tsvector) to find documents that match the query string, ranked by relevance.

Semantic Search

Uses Postgres vector search (pg_vector) to find documents that are semantically similar to the query embedding, ranked by similarity.