Talking Youtube

An AI-powered chat system that allows users to interact with YouTube video content through natural language queries. Built using LangChain and Google's Gemini model.

Features

• Extract transcripts from YouTube videos
• Process and chunk video transcripts
• Generate embeddings using Google's Gemini model
• Create semantic search capabilities using FAISS vector store
• Interactive Q&A with video content
• Multi-query retrieval for better context understanding

Prerequisites

• Python 3.8+
• Google API Key (for Gemini model)
• Virtual Environment (recommended)

Installation

1. Clone the repository:

git clone <repository-url>
cd TalkingYoutube

2. Create and activate a virtual environment:

python -m venv venv
# On Windows
.\venv\Scripts\activate
# On Unix or MacOS
source venv/bin/activate

3. Install required packages:

pip install -r requirements.txt

4. Create a .env file in the root directory and add your Google API key:

GOOGLE_API_KEY=your_api_key_here

Project Structure

TalkingYoutube/
├── assistant/
│   └── prompt.py           # AI assistant prompt templates
├── indexing/
│   ├── document_load.py    # YouTube transcript fetching
│   ├── embedder.py         # Document embedding generation
│   └── splitter.py         # Text splitting utilities
├── utils/
│   ├── formatter.py        # Context formatting
│   └── parser.py          # Output parsing
├── .env                    # Environment variables
├── main.py                # Main application
└── requirements.txt       # Project dependencies

Usage

1. Run the main script with a YouTube video ID:

python main.py

2. The system will:
   • Fetch the video transcript
   • Split it into manageable chunks
   • Generate embeddings
   • Create a vector store
   • Allow you to ask questions about the video content

Example

question = "What is vectorization?"
answer = ask.invoke(question)
print(answer)

Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Open a Pull Request

Acknowledgments

• LangChain for the framework
• Google's Gemini model for embeddings and chat
• FAISS for vector storage
• YouTube Transcript API for caption extraction