VDAOvideo

I Built an Offline AI That Reads Your PDFs and the Pictures in Your Slides

Explain AI · 25 views · 3 weeks ago
Description

Learn how to build a completely offline AI chatbot that answers questions from your own local database — no internet, no ChatGPT or Claude subscription, and no uploading your files to the cloud. In this video I walk through a Retrieval Augmented Generation (RAG) setup using Ollama, a local LLM (Llama 3.2) and a local embedding model to turn your PDFs, Word documents, Excel sheets and PowerPoint files into vector embeddings you can chat with. I demonstrate querying a 900-page software engineering textbook with the Wi-Fi switched off and getting the exact page number, then ask about a product image buried inside slide 7 of a PowerPoint deck and the offline document chat reads the picture and answers correctly. I built this by applying the viral “Reverse Engineer Anything” GitHub repo to the