If you’re looking for an AI-powered solution to interact with your PostgreSQL database using natural language, there are a few interesting options:
pg_gpt (Experimental Extension):
- Description: An experimental extension that brings the OpenAI GPT API to your PostgreSQL database. It allows you to run queries in human language.
- How It Works: The extension sends a subset of your database schema to ChatGPT and asks it to generate a query based on this schema and user input.
Installation:
- First, install pgx using: cargo install --locked cargo-pgx.
- Clone the pg_gpt repository and follow the setup instructions.
- Set your OpenAI API key and create the extension: create extension pg_gpt; set openai.key = '<YOUR OPENAI API KEY HERE>';
Limitations:
- Schema Size: Currently uses gpt-3.5-turbo, which is limited to 4096 tokens. Use gpt_tables to narrow down the set of tables.
Note: This plugin sends schema (without data) to the OpenAI GPT API, so it’s not recommended for production databases1.
pgvector:
- Description: An open-source library that can turn your Postgres DB into a vector database.
- Usage: You can store embeddings using the pgvector extension.
- Availability: Free and open-source2.
postgres-gpt:
- Description: Combines pgvector and OpenAI GPT for PostgreSQL.
Installation:
Use Docker: docker run --name postgres-gpt -p 5432:5432 -e OPENAI_API_KEY=<your_openai_api_key> -e POSTGRES_PASSWORD=postgres -d yulinying/postgres-gpt
Alternatively, use the yulinying/postgres-gpt-playground image with pre-embedded sample document data3
0 Comments