NuExtract: An LLM that extracts information

[@paolo-e-basta]

Loving the uv CLI command at 1:00

[@YufengCHEN-ng4op]

Thanks man, but I think to really do this task in production environment we need some sort of safety net - I mean if the model spit out half-descent JSON, or only partially abide our given schema, What can we do to coerce it back to legit JSON in post-processing?

[@tecnom7133]

Thanks man, I was just thinking about that when i found your video
I have a question : if you know any local document parsers to transform like Invoices into Markdown which is better for an LLM ?

[@anisdouadi6733]

Hello, maybe PyMuPDF