title	app_file	sdk	sdk_version
YoutubeRAG	main.py	gradio	5.46.0

YouTube RAG System

Languages: English · 中文

A lightweight, modular YouTube video Q&A tool with persistent session management. Supports automatic subtitle extraction or audio transcription to build a searchable knowledge base.

Web Interface: Interactive Q&A with session management

✨ New in v2.0

• 🏗️ Modular Architecture: Clean, maintainable code structure
• 💾 Session Persistence: Save and reload RAG sessions
• 🌐 Web Interface: Beautiful Gradio-based UI
• 📱 Simplified CLI: Streamlined command-line experience
• 🔧 Better Error Handling: More robust and user-friendly

Table of Contents

• Features
• Installation
• Quick Start
• Usage
• Project Structure
• License

Features

Core Features

• 🎥 Smart Content Extraction: Automatic YouTube subtitle extraction with audio transcription fallback
• 📝 AI Summarization: Generate concise video summaries using OpenAI models
• 🔍 Vector Search: Intelligent Q&A using RAG (Retrieval Augmented Generation)
• 💾 Session Persistence: Save and reload analysis sessions
• 🌐 Multi-language Support: Chinese and English interfaces

Interface Options

• 🖥️ Web Interface: User-friendly Gradio web UI
• ⌨️ Command Line: Streamlined CLI for developers
• 🔄 Session Management: List, load, save, and delete sessions

Technical Features

• 🏗️ Modular Design: Clean separation of concerns
• 🤖 Multiple AI Models: Support for GPT-3.5, GPT-4, GPT-4o series
• 📊 Chunked Processing: Handle long videos efficiently
• 🛡️ Error Resilience: Robust error handling and recovery

Installation

Prerequisites

• Python 3.12+
• OpenAI API key

Install Dependencies

pip install -r requirements.txt

Set Environment Variables

export OPENAI_API_KEY="your-openai-api-key"

Quick Start

🌐 Web Interface (Recommended)

# Launch web interface (default mode)
python main.py

# Or explicitly specify UI mode
python main.py --ui

Open your browser and navigate to http://localhost:7860

⌨️ Command Line Interface

# Analyze a YouTube video
python main.py --url "https://www.youtube.com/watch?v=VIDEO_ID"

# Use different AI model
python main.py --url "https://www.youtube.com/watch?v=VIDEO_ID" --model gpt-4

# Skip Whisper transcription fallback (only use subtitles)
python main.py --url "https://www.youtube.com/watch?v=VIDEO_ID" --no-transcription

# Custom chunking parameters
python main.py --url "https://www.youtube.com/watch?v=VIDEO_ID" --chunk-size 1500 --chunk-overlap 50

Usage

Web Interface

1. Start the application: python main.py
2. Enter API Key: Provide your OpenAI API key
3. Choose Action:
   • New Video: Analyze a fresh YouTube video
   • Load Session: Continue with a previously saved session
4. Ask Questions: Interact with the AI about the video content

Session Management

# List all saved sessions
python main.py --list-sessions

# Load a specific session
python main.py --load-session "session_name"

# Delete a session
python main.py --delete-session "session_name"

# Generate a detailed analysis report for a saved session
python main.py --export-analysis "session_name" --analysis-language zh

# Ingest every video from a YouTube playlist into one session
python scripts/playlist_ingest.py "https://www.youtube.com/playlist?list=PLAYLIST_ID" \
  --session-name buffet_all --skip-existing --save-summary

Advanced Usage

Custom Model Configuration

# Use GPT-4 for better quality (slower, more expensive)
python main.py --url "VIDEO_URL" --model gpt-4

# Use GPT-4o-mini for faster processing
python main.py --url "VIDEO_URL" --model gpt-4o-mini

Session Management Commands (in Web/CLI interface)

• sessions - View all saved sessions
• save as [name] - Save current session with custom name
• reset - Restart the application
• exit - Quit the application

Project Structure

YouTube RAG System/
├── main.py                          # Main entry point
├── requirements.txt                 # Python dependencies
├── README.md                       # This file
├── README.zh-CN.md                # Chinese documentation
├── src/                           # Source code
│   └── youtube_rag_system/        # Main package
│       ├── __init__.py            # Package initialization
│       ├── core/                  # Core functionality
│       │   ├── __init__.py
│       │   ├── rag_engine.py      # Main RAG engine
│       │   ├── content_processor.py # Video processing
│       │   └── session_manager.py  # Session persistence
│       ├── ui/                    # User interfaces
│       │   ├── __init__.py
│       │   └── gradio_interface.py # Web interface
│       └── utils/                 # Utility functions
│           ├── __init__.py
│           ├── validators.py      # Input validation
│           └── file_utils.py      # File operations
├── rag_sessions/                  # Saved sessions (auto-created)
│   └── [session_id]/
│       ├── metadata.json         # Session metadata
│       └── chroma_db/            # Vector database
└── images/                       # Documentation images
    ├── Demo.png
    └── Demo2.png

How It Works

1. Content Acquisition:

   • First attempts to extract YouTube auto-generated subtitles
   • Falls back to audio download and Whisper transcription if needed

2. AI Processing:

   • Generates comprehensive video summary
   • Splits content into optimized chunks for vector search

3. Knowledge Base Creation:

   • Creates vector embeddings using OpenAI embeddings
   • Stores in ChromaDB for efficient retrieval

4. Session Persistence:

   • Saves all processed data to disk
   • Enables quick reload without reprocessing

5. Interactive Q&A:

   • Uses RAG to answer questions based on video content
   • Maintains context and provides relevant responses

Command Line Options

python main.py [-h] [--ui] [--url URL] [--model MODEL] 
                   [--chunk-size CHUNK_SIZE] [--chunk-overlap CHUNK_OVERLAP] 
                   [--list-sessions] [--load-session SESSION] 
                   [--delete-session SESSION]

YouTube RAG System - YouTube视频RAG问答工具

optional arguments:
  -h, --help            show help message and exit
  --ui                  Launch Gradio web interface
  --url URL             YouTube video URL for CLI mode
  --model MODEL         OpenAI model name (default: gpt-3.5-turbo)
  --chunk-size SIZE     Text chunk size (default: 1000)
  --chunk-overlap SIZE  Text chunk overlap (default: 20)
  --list-sessions       List all saved sessions
  --load-session NAME   Load a saved session by name
  --delete-session NAME Delete a saved session by name

Environment Variables

• OPENAI_API_KEY: Your OpenAI API key (required)

Requirements

• Python 3.12+
• OpenAI API key
• Internet connection (for YouTube access and OpenAI API)

Dependencies

See requirements.txt for the complete list of Python packages.

License

This project is licensed under the MIT License.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Support

If you encounter any issues or have questions, please open an issue on GitHub.