run-llama/llama_index ? reverse-engineered prompt

Build me an open source Python framework for building AI apps that can read documents, extract information, and answer questions from your own data.

I want a clean core library plus optional integrations, so people can start fast with a simple install or add only the pieces they need for their LLM, embedding, and vector database provider. It should support ingesting files, parsing messy documents, indexing content, and doing retrieval based question answering. I also want a document agent style layer that can handle OCR and structured extraction, with enough hooks to plug in different backends later.

Please make it feel polished for real users, with good docs, examples, and a clear way to extend it. If you need to look up current docs online while building, go for it. Also include tests and basic developer tooling so it’s easy to maintain.