A playground with the visualization of function calls

I’ve been working on Knit for some time, it’s a prompt management tool for both individuals and teams.

I just added a new prompt editor (similar as OpenAI’s playground) with the support of the function calls. You can:

  • See the exact function calls (name/arguments) the model returns.
  • Add multiple functions (with json schema).
  • Setup mocked values for functions, so Knit will auto reply to the model with the mocked value, and you can see the final result. (this one’s super useful to replay cases)

It’s really simple to use and straight forward. Try it on Knit, it’s free for everyone. And lmk your thoughts!


6 Likes

This is useful, thanks. The OpenAI playground has the ability to get Python/Node.js code for the current request. LangChain would be great as well. Thinking of adding this?

Thanks for the suggestion and yes, it’s on the roadmap, will show the code for the current request.