https://github.com/bimalkshah/youtube_rag
Youtube Video Bot
https://github.com/bimalkshah/youtube_rag
llm ollama ollama-app rag-chatbot youtube youtubebot
Last synced: 4 months ago
JSON representation
Youtube Video Bot
- Host: GitHub
- URL: https://github.com/bimalkshah/youtube_rag
- Owner: bimalkshah
- License: mit
- Created: 2025-05-05T15:43:52.000Z (5 months ago)
- Default Branch: main
- Last Pushed: 2025-05-05T16:16:59.000Z (5 months ago)
- Last Synced: 2025-05-29T15:11:45.600Z (5 months ago)
- Topics: llm, ollama, ollama-app, rag-chatbot, youtube, youtubebot
- Language: Python
- Homepage:
- Size: 7.81 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# youtube_rag
This project allows you to ask questions about the content of any YouTube video using its transcript. It uses:
1. LangChain for building the pipeline
2. FAISS for semantic search over transcript chunks
3. Ollama to run local LLMs (like TinyLlama)
4. YouTube Transcript API for fetching transcripts
## Features
1. Extracts transcript from any public YouTube video
2. Splits and stores transcript using FAISS vector store
3. Embeds using mxbai-embed-large via Ollama
4. Answers using tinyllama (or other local LLMs)
5. Answers only based on the transcript context
🛠️ Installation
1. Clone the repository
```
git clone https://github.com/your-username/youtube-qa-langchain.git
cd youtube-qa-langchain
```
2. Create and activate virtual environment
```
python -m venv venv
source venv/bin/activate # On Windows use `venv\Scripts\activate`
```
3. Install dependencies
```
pip install -r requirements.txt
```
4. Make sure Ollama is installed and running locally:
```
Install Ollama
```
5. Pull the required models
```
ollama pull mxbai-embed-large
ollama pull tinyllama
```
# Usage
Run the script:
```
python app.py
```
You'll be prompted to:
```
Enter a YouTube Video ID (e.g., dQw4w9WgXcQ)
Ask a question related to the video
```
# Example
```
Please enter Youtube Video ID: VbZ9_C4-Qbo
Ask your question related to the video: What is the speaker's main argument?
Answer: The speaker argues that AI can be aligned through human feedback, but highlights challenges in value alignment.
```