🧬 Self-Evolving AI Agent

A multi-agent app built on EvoAgentX that turns a single natural-language goal into a working program. It automatically generates a multi-agent workflow, executes it to produce code, then verifies and repairs that code with a second model — no manual agent wiring required.

The included example takes the goal "Generate HTML code for a Tetris game that can be played in the browser" and writes a ready-to-play index.html.

✨ What It Demonstrates

• Automatic workflow generation — WorkFlowGenerator designs the agents and steps from a plain-English goal.
• Multi-agent execution — AgentManager + WorkFlow instantiate and run the generated agents.
• Cross-model code verification — generation runs on OpenAI gpt-4o-mini; a separate Anthropic Claude pass verifies and fixes the output.
• Self-evolving by design — the workflow is built and refined by the system itself rather than hand-coded.

🛠️ How to Get Started

git clone https://github.com/Shubhamsaboo/awesome-llm-apps.git
cd awesome-llm-apps/advanced_ai_agents/multi_agent_apps/ai_self_evolving_agent

2. Install dependencies

   pip install -r requirements.txt
   pip install git+https://github.com/ANative-Lab/EvoAgentX.git

3. Set your API keys