Odeon

AI Agent Optimization Platform

Simulate, Evaluate, and Evolve AI Agents with Autonomous Feedback Loops.

---

📖 Table of Contents

• Overview
• Key Features
• System Architecture
• Technology Stack
• Quick Start
• Configuration
• API Reference
• Roadmap
• Troubleshooting
• Contributing
• License
• Contact

---

🔭 Overview

Odeon is a cutting-edge playground for AI Agent Engineering. It solves the "black box" problem of prompt tuning by automating the evaluation loop. Instead of manually tweaking prompts and hoping for better results, Odeon:

1. Simulates realistic user interactions (e.g., a stubborn debt defaulter).
2. Evaluates the agent's performance against strict numerical KPIs (Empathy, Negotiation, Repetition).
3. Optimizes the system prompt automatically using a meta-agent if targets are missed.

The result is a self-improving agent that converges on the optimal persona for your specific business goals.

---

🚀 Key Features

• ⚡ Autonomous Optimization Loop

  • Generates diverse user personas (e.g., "The Lawyer", "The Crying Student").
  • Runs high-fidelity simulations using Groq for near-instant inference.
  • Rewrites prompts automatically based on granular feedback.

• 📡 Real-Time Simulation Stream

  • Bi-directional WebSocket integration.
  • Watch agent interactions unfold character-by-character.
  • Live state tracking of current optimization cycles.

• 🔄 Neural Visual Diffing

  • Git-style Red/Green diff viewer for Prompt Evolution.
  • See exactly which words changed to improve empathy or compliance.

• 🎨 Neo-Brutalist / Glassmorphism UI

  • A high-end, distraction-free interface built with Tailwind CSS 4.
  • Dark mode focused "Deep Space" aesthetic.

• 📊 Strict Metric Thresholds

  • Define pass/fail criteria (1-10) for Repetition, Negotiation, and Empathy.
  • Agents must meet all criteria to "pass" a scenario.

• 🗄️ SQLite History & Replay

  • Every run is archived. You can replay, analyze, and fork past simulations.

---

🏗️ System Architecture

Odeon uses a decoupled, event-driven architecture to handle high-concurrency simulations.

graph TD
    User[User / Browser] -->|HTTP / WS| FE["Frontend (React + Vite)"]
    FE -->|WebSocket| API["Backend API (FastAPI)"]
    
    subgraph "Backend Enclave"
        API -->|Dispatch| Sim[Simulator Engine]
        Sim -->|Generate| Gen[Persona Generator]
        Sim -->|Chat| Agent[Agent LLM]
        Sim -->|Chat| UserSim[User Simulator LLM]
        
        Sim -->|Data| Eval[Evaluator]
        Eval -->|Feedback| Opt[Prompt Optimizer]
        Opt -->|New Prompt| Agent
        
        Agent <--> Groq[Groq Llama 3 API]
        UserSim <--> Groq
        Eval <--> Groq
        Opt <--> Groq
    end
    
    API -->|Read/Write| DB[(SQLite History DB)]

Loading

---

🛠️ Technology Stack

Component	Tech	Description
Backend	Python 3.10+	Core Application Logic
API Framework	FastAPI	Async, High-performance REST & WS
AI Inference	Groq Cloud	Llama 3.1-8b / 70b (Ultra-fast)
Orchestration	LangChain	Chain Management & Parsing
Database	SQLite	Lightweight embedded persistence
Frontend	React 19	UI Library with Concurrent Mode
Build Tool	Vite	Instant HMR & bundling
Styling	Tailwind CSS 4	Utility-first CSS engine
Type Safety	TypeScript	End-to-end typing

---

⚡ Quick Start

Prerequisites

• Python 3.10+
• Node.js 18+ & npm
• Groq API Key (Get it free at console.groq.com)

1. Clone & Install

git clone https://github.com/vasu-devs/odeon.git
cd odeon

2. Backend Setup

cd backend
python -m venv venv

# Activate Venv
source venv/bin/activate  # Mac/Linux
# venv\Scripts\activate   # Windows

pip install -r requirements.txt

3. Frontend Setup

cd ../frontend
npm install

---

⚙️ Configuration

Create a .env file in the backend/ directory:

# Required: The engine power
GROQ_API_KEY=gsk_your_key_here

# Optional: For experimental multi-model support
GEMINI_API_KEY=your_gemini_key

Running the App

Terminal 1 (Backend):

cd backend
# Make sure venv is active
python server.py

Terminal 2 (Frontend):

cd frontend
npm run dev

Visit http://localhost:5173 to launch Odeon.

---

🔌 API Reference

WebSocket Protocol (/ws/simulate)

Request (Start Simulation):

{
  "api_key": "gsk_...",
  "model_name": "llama3-8b-8192",
  "base_prompt": "You are a specialized agent...",
  "thresholds": { "negotiation": 8.0, "empathy": 7.5 }
}

Response (Events):

• log: Raw system output.
• result: Final conversation metrics.
• optimization: Diff of the prompt change.

---

🔮 Roadmap

• Multi-Agent Swarms: Simulating group dynamics.
• Vector Memory: Giving the agent long-term memory across runs.
• Cloud Deploy: One-click deploy to Vercel/Railway.
• Custom Models: Support for Anthropic/OpenAI via LiteLLM.
• Export Results: PDF/CSV export for compliance reporting.

---

❓ Troubleshooting

Q: I get a 401 Unauthorized error from Groq. A: Check your .env file. Ensure GROQ_API_KEY is set correctly and has no trailing spaces.

Q: The frontend shows "Disconnected". A: Ensure the backend is running on port 8000. Check the terminal for any Python traceback errors.

Q: Optimize Loop isn't updating the prompt. A: Ensure your "Overall" threshold isn't set too low. If the agent passes the low threshold, it won't optimize. Increase the target scores.

---

🤝 Contributing

Contributions are what make the open-source community such an amazing place to learn, inspire, and create. Any contributions you make are greatly appreciated.

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your Changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

---

📄 License

Distributed under the MIT License. See LICENSE for more information.

---

✍️ Authors

• Vasudev Siddh - Initial Work - vasu-devs

Built with ❤️ by the Vasu-devs