docling-project/docling-serve — reverse-engineered prompt

Build me a small service that runs Docling as a web API so I can send it documents or document URLs and get back converted content through a stable v1 endpoint. I want it to be easy to start locally with Python, and also easy to run in a container. Please include a simple browser UI playground at a /ui route so I can test conversions without writing code, plus normal API docs in the browser.

Make the default local server run on port 5001, with clear routes for the main API, docs, and UI. It should support a basic convert request like giving it an http source URL and getting JSON back. Please wire in sensible runtime configuration with environment variables, and keep the setup friendly for both CPU only and optional GPU container use where available.

If there are already docs or examples in the repo, follow those patterns and fill in anything missing so it feels complete and usable. Look up current docs online if you need to.