If you ask a plain chatbot about your own PDF, it either can’t see the file or it guesses. Even when it can read the file, you can’t tell whether the answer came from your document or from the model’s memory. OpenResearch fixes this by making the model answer only from passages pulled out of your documents, and by showing those passages next to the answer.

The technique is called RAG (Retrieval-Augmented Generation). The model doesn’t memorize your documents. At question time, the app first looks up the relevant passages, then hands them to the model and says “answer using only these.”