joelhooks/pdf-brain ? reverse-engineered prompt

Build me a local first tool called pdf brain that lets me create a personal knowledge base from PDFs and Markdown files. I want to be able to add single files, whole folders, or even URLs, then search everything by meaning, by plain text, or with a mix of both. Keep it private and runnable on my own machine, using Ollama for embeddings and optional AI enrichment so each document can get a cleaned up title, summary, tags, category, and suggested concepts.

Make it feel like a solid command line app with commands to init the library, check status, add, ingest, search, list, read, remove, tag, and show stats. It should support a simple concept taxonomy with a tree, search, add, seed, and assign flow. I also want it to work nicely with coding assistants through an MCP server, and default output should be machine readable JSON with a human readable text option too. If you need details, look up current docs online and make sensible choices.