wtf is youtube-rag-chatbot-?

This project lets you chat with the content of any YouTube video in plain language, instead of asking a general purpose chatbot that only knows what it was trained on. You paste in a YouTube link, and the app pulls the video's transcript, breaks it into smaller pieces, and turns those pieces into embeddings, which are numerical representations that let a computer compare meaning between pieces of text. It stores those embeddings in a FAISS vector database, a type of search index built for finding similar text quickly. A key design choice here is that every video gets its own separate vector database, named after the video's unique YouTube ID. This means when you ask a question, the chatbot only searches within that one video's transcript, so it will not accidentally pull in unrelated information from a different video. If a video has already been processed once, the app reuses the saved database instead of redoing the work, which saves time on repeat visits. When you ask a question, the app finds the most relevant chunks of the transcript, builds a prompt with that context, and sends it to a locally running Llama 3.1 8B model through Ollama to generate an answer grounded in what the video actually says. The backend is built with FastAPI and stores user accounts, videos, and conversations in a SQLite database. The frontend is plain HTML, CSS, and JavaScript, and the whole system supports multiple users with login accounts. The author describes this as a learning project covering retrieval augmented generation, vector search, and full stack AI development, and lists planned improvements such as stronger authentication, persistent chat history, support for more file types, and cloud deployment. Suggested uses include building an assistant for podcasts, lectures, corporate training videos, or internal documentation. The README does not mention a license.