postgres-semantic-search

What problem does it solve?

This skill enables PostgreSQL-based semantic and hybrid search using pgvector and ParadeDB.

Core Features & Use Cases

• Semantic search: leverage pgvector to retrieve meaning-based results from documents stored in PostgreSQL.
• Hybrid search: combine full-text search with vector similarity (BM25/RRF) for robust ranking, including ParadeDB as an Elasticsearch alternative.
• Scalability: supports index types (HNSW, IVFFlat) and large datasets, enabling efficient search in product catalogs, documents, and knowledge bases.
• Use Case: build a knowledge-base search for a software team that returns contextually relevant articles and code references.

Quick Start

Begin by enabling pgvector in PostgreSQL, optionally install ParadeDB, create the embedding column, and run the included SQL scripts to set up semantic search functions and indexes. Use the provided embeddings script to generate embeddings with OpenAI and store them in your vector columns, then run semantic or hybrid searches against your data.