A terminal-based CLI system for running various specialized AI agents powered by Google Gemini and File Search.
- Menu-Driven Interface: Beautiful terminal UI with color-coded output
- Multiple Specialized Agents: Extensible architecture for different task-specific agents
- Gemini File Search Integration: Full RAG capabilities with automatic citations
- Agent Configuration: Each agent has unique personality, prompts, and knowledge bases
Downloads and analyzes YouTube channel content:
- Complete video transcripts
- Video metadata (titles, descriptions, views, likes)
- Comments and replies with engagement metrics
- Upload dates and statistics
Then chat with the content using natural language:
- "What topics does this creator focus on?"
- "Show me the most popular videos"
- "What are viewers saying about [topic]?"
- Get answers with automatic citations from File Search!
- Python 3.9 or higher
- Google Gemini API key
- YouTube Data API v3 key
-
Clone the repository
git clone <repository-url> cd cli
-
Install dependencies
pip install -r requirements.txt
-
Set up API keys
Gemini API Key:
- Visit https://ai.google.dev/
- Create an account or sign in
- Go to "Get API Key"
- Create a new API key
YouTube API Key:
- Go to https://console.cloud.google.com/
- Create a new project or select existing
- Enable "YouTube Data API v3"
- Create credentials (API Key)
Set environment variables:
export GEMINI_API_KEY='your-gemini-api-key' export YOUTUBE_API_KEY='your-youtube-api-key'
To make them permanent, add to your
~/.bashrcor~/.zshrc:echo "export GEMINI_API_KEY='your-gemini-api-key'" >> ~/.bashrc echo "export YOUTUBE_API_KEY='your-youtube-api-key'" >> ~/.bashrc source ~/.bashrc
python main.py-
Download Channel Data
- Select "YouTube Channel Analyzer" from main menu
- Choose "Download channel data"
- Enter YouTube channel URL or ID
- Wait for download to complete
- Choose whether to upload to File Search
-
Chat with Content
- Select "Chat with channel content"
- Choose a downloaded channel
- Start asking questions!
You: What are the main topics this creator covers?
You: Which video has the most engagement?
You: What do viewers think about [specific topic]?
You: Summarize the content from the last 5 videos
You: Are there any common complaints in the comments?
cli/
├── main.py # Main CLI entry point
├── agents/
│ ├── __init__.py
│ ├── base.py # Base agent classes
│ └── youtube_agent.py # YouTube analyzer agent
├── utils/
│ ├── __init__.py
│ ├── ui.py # Terminal UI utilities
│ └── gemini_file_search.py # Gemini File Search wrapper
├── config/ # Agent configurations
├── data/ # Agent knowledge bases
│ └── youtube_analyzer/ # YouTube agent data
│ └── {channel_name}/ # Per-channel directories
│ ├── channel_info.json
│ ├── complete_data.json
│ ├── video_*.json
│ └── .file_search_store # File Search store ID
├── logs/ # Application logs
├── requirements.txt # Python dependencies
├── CLAUDE.md # Claude instructions
└── README.md # This file
To add a new agent:
-
Create agent class in
agents/your_agent.py:from agents.base import BaseAgent, AgentConfig from utils.ui import UI class YourAgent(BaseAgent): def __init__(self): config = AgentConfig( id="your_agent", name="Your Agent Name", description="What your agent does", personality="Agent's personality/style", system_prompt="Agent's system instructions", model="gemini-1.5-pro" ) super().__init__(config) def run(self, ui: UI): # Implement your agent logic pass
-
Register in main.py:
from agents.your_agent import YourAgent # In main(): registry.register(YourAgent())
This system uses Google's File Search feature, which provides:
- Automatic RAG: Handles chunking, embeddings, and retrieval
- Vector Search: Semantic understanding of queries
- Built-in Citations: Responses cite sources automatically
- Multi-format Support: PDF, DOCX, TXT, JSON, code files
- Indexing: $0.15 per 1M tokens (one-time per file)
- Storage: Free (1GB on free tier)
- Query-time Search: FREE
The project includes a comprehensive test suite to validate all components:
python3 test_suite.py- Module Imports: All components import correctly
- UI Components: Terminal UI and color system
- Agent System: Base classes, registry, configuration
- YouTube Agent: Initialization, methods, data handling
- File Search: Manager initialization
- Data Storage: Save/load operations
- Error Handling: Edge cases and missing data
All tests run without requiring API keys - they validate code structure and logic.
✅ All 18 tests passed!
Error: Missing API keys: GEMINI_API_KEY
→ Make sure you've exported the environment variables
YouTube API has daily quota limits. If you hit the limit:
- Wait 24 hours for reset
- Or create additional API keys in Google Cloud Console
ModuleNotFoundError: No module named 'google.generativeai'
→ Run pip install -r requirements.txt
PermissionError: [Errno 13] Permission denied: 'data/'
→ Make sure the CLI has write permissions in the directory
- Web Scraper Agent
- UI Builder Agent
- Code Analysis Agent
- MCP Server Integration
- Custom tool support per agent
- Export chat histories
- Multi-language support
Feel free to add new agents or improve existing ones!
MIT License
For issues or questions:
- Check the Troubleshooting section
- Review the agent code in
agents/ - Check Gemini API docs: https://ai.google.dev/gemini-api/docs
- Check YouTube API docs: https://developers.google.com/youtube/v3