llmware-ai/llmware ? reverse-engineered prompt

Build me a Python framework that makes it easy to create AI applications that can reason over my own private documents. The main feature I want is the ability to create a "knowledge base" by just pointing the tool at a folder full of files—it should automatically handle things like PDFs, Word docs, and PowerPoints. Then, I need to be able to ask questions and have an LLM use the information from those specific documents to generate an answer.

A key part of this should be a large, built-in catalog of models, especially small and efficient ones that can run locally on my laptop for privacy and speed. It should be super simple to load and use any of these models, whether they're local or from a cloud API like OpenAI, using a single, easy interface. Essentially, I'm looking for a complete, unified toolkit for building secure, local RAG pipelines from scratch.