Exploring declarative AI with Python — would love your thoughts

[AlbanPerli]

Hi everyone,

I’ve been experimenting with a small library that allows users to build functions or programs for LLMs by creating a kind of “programmable” prompt in Python:
https://github.com/AlbanPerli/Noema-Declarative-AI

I may be misunderstanding some things, but it feels like not many people see the usefulness of this kind of approach. I’d really appreciate any thoughts, feedback, or advice from those with more experience.

Thanks for taking the time to read, and I look forward to learning from the discussion.

[vazirim]

Hi @AlbanPerli, thanks for the pointer, this looks interesting. How is the evaluation implemented? Is that with an llm-as-judge, or is the same model evaluating itself?

Would you be interested to contribute a PDL front-end to Noema?