Postgres without SQL: Natural language queries using GPT-3 & Rust
summary
A Rust Postgres extension called pg_human grabs the live schema, sends it to GPT‑3/ChatGPT via an OpenAI Rust client, receives raw SQL, runs it through Postgres SPI and returns JSON for arbitrary results. Demo creates todo tables, inserts Messi/Ronaldo/Neymar with three tasks each, updates ~50% of tasks, and runs GROUP BY queries successfully. Notes cover using cargo‑pgx/pgrx, async/threading limits in Postgres, a mold linker speedup, and a warning about destructive AI-generated ALTER/DELETE.
AI-written summary · about
chapters (7)
- 0:00Introducing Jelte
- 1:06Jelte’s talk begins: Querying Postgrwith ChatGPT
- 5:11Why do this in Rust?
- 8:52Let’s write a Postgres extension in Rust
- 13:31Requirements for a PG extension for ChatGPT
- 19:25Demo of a To-Do list app
- 26:43Q&A
related talks
26mCan We Use Rust to Develop Extensions for PostgreSQL?
53mModern PostgreSQL for AI...
39mDeveloping a Postgres Extension in Rust
46mFearless Extension Development With Rust and PGRX
11mPostgres plays Pokémon!
37mRound Table Discussion
01h 00mPostgreSQL and Artificial intelligence
50mLeveraging AI as a PostgreSQL DBA