pg_vectorize is a Postgres extension that simplifies text transformation to embeddings and integrates with popular LLMs for vector search and LLM applications with just two function calls. It relies on pgvector, pgmq, and SentenceTransformers. Features include workflows for vector search and RAG, integration with OpenAI’s embeddings and chat-completion endpoints, and automated creation of triggers to update embeddings. The setup process involves running the Tembo docker container and vector server, creating a job to vectorize data, and conducting searches. Real-time trigger updates keep embeddings current. With support for Hugging Face Sentence-Transformers, pg_vectorize offers efficient data processing capabilities within Postgres.
https://github.com/tembo-io/pg_vectorize