Query
  |
[Generate embedding] --> Vector Search (PGVector) --> Top 30 results
  |
[Generate ts_query]  --> Keyword Search (BM25)    --> Top 30 results
  |
[Reciprocal Rank Fusion (RRF)] --> Merge & re-rank --> Top 10 final results

rrf_score = 1/(k + vector_rank) + 1/(k + keyword_rank)  # k=60 (standard)

CREATE TABLE chunks (
    id UUID PRIMARY KEY,
    document_id UUID REFERENCES documents(id),
    content TEXT NOT NULL,
    embedding vector(1024),  -- PGVector
    content_tsvector tsvector GENERATED ALWAYS AS (
        to_tsvector('english', content)
    ) STORED,
    section_title TEXT,
    content_type TEXT,
    created_at TIMESTAMP DEFAULT NOW()
);

-- Indexes
CREATE INDEX idx_chunks_embedding ON chunks
    USING hnsw (embedding vector_cosine_ops)
    WITH (m = 16, ef_construction = 64);

CREATE INDEX idx_chunks_content_tsvector ON chunks
    USING gin (content_tsvector);

async def hybrid_search(
    query: str,
    query_embedding: list[float],
    top_k: int = 10
) -> list[Chunk]:
    FETCH_MULTIPLIER = 3  # Fetch 30 for better RRF coverage
    K = 60  # RRF smoothing constant

    # Vector search subquery
    vector_subq = (
        select(Chunk.id,
            func.row_number().over(
                order_by=Chunk.embedding.cosine_distance(query_embedding)
            ).label("vector_rank"))
        .limit(top_k * FETCH_MULTIPLIER)
        .subquery()
    )

    # Keyword search subquery
    ts_query = func.plainto_tsquery("english", query)
    keyword_subq = (
        select(Chunk.id,
            func.row_number().over(
                order_by=func.ts_rank_cd(Chunk.content_tsvector, ts_query).desc()
            ).label("keyword_rank"))
        .where(Chunk.content_tsvector.op("@@")(ts_query))
        .limit(top_k * FETCH_MULTIPLIER)
        .subquery()
    )

    # RRF fusion with FULL OUTER JOIN
    rrf_subq = (
        select(
            func.coalesce(vector_subq.c.id, keyword_subq.c.id).label("chunk_id"),
            (func.coalesce(1.0 / (K + vector_subq.c.vector_rank), 0.0) +
             func.coalesce(1.0 / (K + keyword_subq.c.keyword_rank), 0.0)
            ).label("rrf_score"))
        .select_from(vector_subq.outerjoin(keyword_subq, ..., full=True))
        .order_by("rrf_score DESC")
        .limit(top_k)
        .subquery()
    )

    return await session.execute(
        select(Chunk).join(rrf_subq, Chunk.id == rrf_subq.c.chunk_id)
    )

results = await hybrid_search(
    query="binary search",
    query_embedding=embedding,
    content_type_filter=["code_block"]
)

results = await hybrid_search(query, embedding, top_k=50)
filtered = [r for r in results if (1 - r.vector_distance) >= 0.75][:10]

queries = ["machine learning", "ML algorithms", "neural networks"]
all_results = [await hybrid_search(q, embed(q)) for q in queries]
final = deduplicate_and_rerank(all_results)

FT.HYBRID index query
  VECTOR $embedding AS vec_score
  TEXT "search terms" AS text_score
  COMBINE RRF
  LIMIT 0 10
  RETURN 3 content title score