FastAPI agent orchestration server with OpenAI-compatible chat endpoints and running short multi-turn debate on Ollama. This example uses gemma3:4b, change in docker-compose.yml.
# Automatic build
./build_containers.sh
# Follow logs
sudo docker compose logs -f agent-orchestrator
# Optional: open dashboard in browser
# http://localhost:11435/dashboard
# Dashboard "Run start_1" shows live orchestrator logs in the HTML console.
# Run the experiment directly inside the container (no HTTP)
sudo docker compose exec agent-orchestrator python main.py --run-trigger start_1
# CLI trigger logs appear only in that terminal session (not in the dashboard stream).
# Optional: open a shell in the container
sudo docker compose exec agent-orchestrator bash