Multi-Model AI Chatbot System

A modular Python-based AI agent system that integrates OpenAI, Anthropic, Google Gemini, and DeepSeek.

Features

• Intelligent Routing: Automatically selects the best model (Coding -> DeepSeek, Reasoning -> Claude, Fast -> Gemini).
• Ensemble Mode: Queries multiple models and uses a judge (GPT-4o) to rank and select the best response.
• Reflection Loop: Self-critique mechanism (Generate -> Critique -> Improve).
• Memory: Maintains conversation history.
• Cost Tracking: Tracks usage costs per model in real-time.
• Rich CLI: Beautiful terminal interface using rich.

Setup

1. Install Dependencies:

   pip install -r requirements.txt

2. Configure Environment Variables: Copy .env.example to .env and fill in your API keys:

   cp .env.example .env

3. Run the Chatbot:

   python -m app.main

CLI Commands

• /model <model_id>: Switch to a specific model (e.g., gpt-4o, claude-3-5-sonnet, gemini-1.5-flash, deepseek-chat).
• /mode <mode>: Switch mode (auto, ensemble, reflection).
• /clear: Clear conversation memory.
• /cost: View total session cost and breakdown.
• /exit: Quit the application.

Project Structure

• app/models/: Wrapper classes for different AI providers.
• app/agents/: Routing and orchestration logic.
• app/memory/: Conversation history management.
• app/utils/: Logging, cost tracking, and prompt building.
• app/config.py: Central configuration and pricing.