ChatGPT plugins now support Postgres & Supabase

Retrieval plugins allow ChatGPT to access information from a database, helping it answer questions from private datasets. Supabase recently contributed to the OpenAI repo with Postgres and Supabase implementations for building plugins using pgvector. A retrieval plugin turns documents into smaller chunks, converts these chunks into embeddings using OpenAI's text-embedding-ada-002 model, stores the embeddings in a vector database, and queries the vector database for relevant documents when a question is asked. Supabase added two vector provider options: one for Postgres and one for Supabase. The main differences are that the Postgres version uses the psycopg2 python library to directly connect to the database, while the Supabase version interacts with the database via PostgREST. Both have the same schema so you can easily switch between them.