NidarshN/local_rag_ollama — reverse-engineered prompt
Reverse engineered prompt
Build me a local command line tool that lets me ask questions about my own PDF files without sending them to the cloud.
I want to drop PDFs into a folder, run one command to index them, then run another command with a question like “What is the main conclusion of this document?” and get a useful answer based only on those PDFs. It should use Ollama locally for the language model, use a local vector database for searching the PDF content, and keep setup simple with Python and uv.
Please include a sample env file, clear install and usage instructions, and commands for both indexing PDFs from the default data folder and from a custom folder path. For now it only needs to support one question at a time from the terminal, not a full chat app. Look up current docs online if you need to.
Want more depth? Deep Reverse