Local AI Document Q&A

About this project

A fully local Retrieval Augmented Generation (RAG) web app. Upload PDFs and text files to a vector library, then ask questions across all documents simultaneously — answered by a local LLM with no data leaving your machine.

What I Learned

• This was my first RAG project that I finished and deployed. A lot of my earlier projects died half-built — so with DocuMind, finishing and shipping it was a personal rule I set for myself before writing a single line of code.
• I learned how to run AI entirely on a local machine using Ollama. No API keys, no cloud, nothing leaves the device. Understanding how a model actually receives a prompt and returns a response made AI feel far less like a black box.
• I built a RAG pipeline from scratch — chunking documents into small pieces of text, converting those chunks into vectors, and passing the relevant ones alongside the user's question to the model. Following that data flow end-to-end was the biggest technical takeaway from this project.