Autonomous Agent (Python Only)

Simple websocket chat server in Python with a single bash tool.

Features

• One Python websocket server (WS /ws)
• One health endpoint (GET /health)
• One agent tool: bash
• OpenRouter model via pydantic-ai
• Works locally and in Docker

Required Inputs

• OPENROUTER_API_KEY (required)
• OPENROUTER_MODEL (optional, default moonshotai/kimi-k2.5)
• WORKSPACE_PATH (optional, default ./workspace locally or /workspace in Docker)

Local Run

python3 -m venv venv
source venv/bin/activate
pip install -r agent/requirements.txt

OPENROUTER_API_KEY="your-key" \
OPENROUTER_MODEL="moonshotai/kimi-k2.5" \
WORKSPACE_PATH="$PWD/workspace" \
bash scripts/start.sh

Docker Run

Build image:

docker build -t autonomous-agent .

Run one agent:

docker run --rm -p 8000:8000 \
  -e OPENROUTER_API_KEY="your-key" \
  -e OPENROUTER_MODEL="moonshotai/kimi-k2.5" \
  -e WORKSPACE_PATH="/workspace" \
  -v "$PWD/workspace:/workspace" \
  autonomous-agent

Run multiple isolated agents:

docker run --rm -p 8001:8000 -e OPENROUTER_API_KEY="your-key" -v "$PWD/ws1:/workspace" autonomous-agent
docker run --rm -p 8002:8000 -e OPENROUTER_API_KEY="your-key" -v "$PWD/ws2:/workspace" autonomous-agent

Each container gets isolated filesystem via a different mounted workspace.

Test Client (Python)

python3 test_client.py --ws-url ws://localhost:8000/ws

Commands:

• normal text: send chat
• /ping: send ping
• /quit: exit