justmichael-80/inbox-ingestor

Inbox Ingestor is a small Python tool that watches a folder on your computer and automatically turns any PDF dropped into it into a clean Markdown file with the same name, placed right next to it. Everything happens locally on your own machine, with no API key, no cloud service, no rate limits, and no cost, using a library called pymupdf4llm to do the actual conversion. The problem it solves is specific: people who pair an AI assistant like Claude or ChatGPT with a note taking system such as Obsidian often find that these assistants struggle to read PDFs directly, either failing outright or asking you to paste the text in by hand. Cloud based conversion services can fix that, but they bring API keys, usage limits, and billing along with them for what should be a simple, repeatable chore. Inbox Ingestor sidesteps that entirely by running the conversion on your own machine. It is intentionally limited in scope. It watches, converts, and stops there, it does not decide where a note belongs in your vault or how it should be categorized, since the author considers that a judgment call better left to an AI assistant with access to your files. It also will not work on scanned or image only PDFs that have no real text layer, since that would require optical character recognition, which this tool does not attempt. Text based PDFs such as most papers, forms, and reports convert well. Setup is a couple of pip installs followed by running one Python script and pointing it at the folder you want watched, and it works the same way on macOS, Windows, and Linux. The README also includes detailed instructions for running it permanently in the background on each operating system, including specific troubleshooting steps for macOS privacy permissions that can silently block a background process from reading your files. The suggested workflow is to drop PDFs into the folder throughout the week, then periodically ask your AI assistant to process the resulting Markdown files and file them properly into your notes. It is released under the MIT license.