asutosh936/rag-demo

Rag-demo lets you ask questions about PDF documents and get answers back in plain language. Instead of scrolling through a long report or set of notes to find a specific detail, you upload the file, type a question, and the tool searches the document for relevant passages and generates a response. It's like having a small research assistant that has read your files and can answer questions about them. Under the hood, the tool breaks each PDF into smaller chunks of text, converts those chunks into a format a computer can search by meaning (called embeddings), and stores them in a local database. When you ask a question, it finds the most relevant chunks and sends them to an AI language model along with your question. The model then produces an answer grounded in your actual documents rather than its general training knowledge. You can use it in two ways: a command-line script where you pass in file paths and a question, or a simple web interface where you upload PDFs in your browser and chat with them. The web UI also caches your uploaded documents during the session, so follow-up questions are faster. This is useful for anyone who works with lengthy documents, a founder reviewing legal contracts, a PM digging through technical specs, or a student analyzing research papers. You need an OpenAI API key to run it, since the final answer generation relies on OpenAI's language model. The README notes that the default text chunking and embedding settings prioritize speed over quality, and you can swap in different models with minimal code changes if you want better results.