ucbepic/docetl ? reverse-engineered prompt

Build me a simple app or toolkit that lets me process a bunch of messy documents and records with LLMs in a more automatic way.

I want to be able to describe tasks in plain English like classifying support tickets, pulling out fields, filtering items, and summarizing groups, then run those steps over lots of input data and get clean table like results back. It should support both a Python workflow and a low code YAML style setup, and it would be great if there’s a command line way to run a pipeline too. Make it easy to try on a small sample first, then scale up to the full dataset.

If it makes sense, include a simple visual playground for editing prompts and seeing results, plus a way to keep track of cost and output shape. Use current docs online if you need to, and make the experience feel polished and easy for someone who just wants to transform documents without wiring every LLM call by hand.