Quick Start
Prerequisites
- Python ≥ 3.8
- At least one LLM API key or local model
1. Install Dependencies
bash
git clone https://github.com/your-repo/MPE_muiltiagent_benchmark.git
cd MPE_muiltiagent_benchmark
pip install -r requirements.txt2. Configure API Key
bash
# Option A: .env file (recommended)
echo 'DEEPSEEK_API_KEY=sk-your-key' > .env
# Option B: Environment variable
export DEEPSEEK_API_KEY="sk-your-key"
# Option C: Runtime parameter
engine = get_api_engine("deepseek", api_key="sk-your-key")3. Run Your First Game
bash
python simple.py # Simplest environment
python spread_API.py # Cooperative coverage
python tag_API.py # Predator-prey chase4. View Output
Each run generates:
xxx_demo.mp4— Game replay videoxxx_demo.json— Step-by-step log (obs, action, thought, reward)
JSON Log Sample
json
{
"step": 5,
"agent": "agent_0",
"observation": { "vel": [0.1, -0.2], "landmark_rel": [-0.3, 0.5] },
"action": [0.0, 0.8, 0.0, 0.0, 0.5],
"thought": "Target is left and above, applying thrust",
"reward": -0.34
}Next Steps
- 🎮 Game Environments — All 9 environments explained
- 🧠 Prompt Engineering — Prompt system design
- 📊 Benchmark — Batch evaluation guide