What are partial indexes in pgvector and when should you use them?

-- Partial index: only index documents in a specific category
CREATE INDEX ON documents USING hnsw (embedding vector_cosine_ops)
WHERE category = 'technical';  -- index only 'technical' rows

-- Queries using the same filter benefit from the smaller index:
SELECT id, content FROM documents
WHERE category = 'technical'                  -- matches the partial index
ORDER BY embedding <=> '[...]' LIMIT 5;      -- uses the partial index
-- Much faster: index is smaller and focused

-- Partial index for published articles only:
CREATE INDEX ON articles USING hnsw (embedding vector_cosine_ops)
WHERE is_published = TRUE;

-- Partial index to exclude NULL embeddings:
CREATE INDEX ON documents USING hnsw (embedding vector_cosine_ops)
WHERE embedding IS NOT NULL;
-- Useful if some rows have no embedding yet

-- Multiple partial indexes for different categories:
CREATE INDEX idx_tech  ON docs USING hnsw (embedding vector_cosine_ops) WHERE category = 'tech';
CREATE INDEX idx_legal ON docs USING hnsw (embedding vector_cosine_ops) WHERE category = 'legal';
-- Each index is smaller and faster for its category

-- PostgreSQL uses a partial index only when the query WHERE clause
-- includes the index's WHERE condition