YT Bot using RAG
Brain Tumor Classification🎥 YouTube RAG Assistant
A Streamlit-based app that lets you ask questions or get a summary of any YouTube video using its transcript, powered by RAG (Retrieval-Augmented Generation).
🚀 Features
- 🔗 Paste a YouTube URL and automatically extract the transcript
- 🌐 Supports multi-language transcripts (auto-translates to English for Q&A)
- 🤖 Choose between:
- OpenAI (GPT)
- Gemini (Google)
- Ollama (local LLMs like llama3)
- 💬 Ask custom questions or auto-summarize the entire video
🔧 Setup Instructions
1. Clone the repo & install requirements
git clone https://github.com/Chaganti-Reddy/YT_RAG.git
cd YT_RAG
pip install -r requirements.txt
2. Run the app
streamlit run app.py
Model Support
Currently, only the following models are supported:
| Model | Needs API Key | Notes |
|---|---|---|
| OpenAI | ✅ Yes | Use GPT-3.5 or GPT-4 |
| Gemini | ✅ Yes | Uses raw API (faster than LangChain wrapper) |
| Ollama | ❌ No | Requires ollama installed locally & model pulled (ollama run llama3) |
Folder Structure
YT_RAG/
├── app.py ← Streamlit UI
├── query_engine.py ← QA chain logic
├── model_selector.py ← Model switcher
├── retriever_utils.py ← Vector store builder
├── transcript_loader.py ← Transcript & translator
├── gemini_direct.py ← Direct Gemini API call
├── requirements.txt
└── README.md
Tips
- Your questions must be in English (auto translation of transcript is handled).
- Gemini API quotas are low on free tier — fallback to Ollama if needed.
- Works great with videos that have auto-generated subtitles.
