We made an LLM Playground where you can use jinja to render prompts and extract structured data. For example, receipts, invoices, audio snippets, etc. It uses the BAML programming language under the hood.
I was tired of not being able to test every model (local + hosted) in the same place. I wanted it work for audio and for images and for text. Plus, I HATE navigating UIs when I can just write code, so the existing non-code first offerings sucked for me.
We were inspired a little bit by Markdown Preview and a little bit by Rust.
For example, the VSCode version of this has a run-tests buttons right in your IDE cause that is the only reason i started writing tests in rust.
What do you think?
Code is here: https://github.com/BoundaryML/baml/tree/canary/typescript
No comments yet.
