Skip to content

See docs/DISCLAIMER_SNIPPET.md This repository is a conceptual research prototype. References to "AGI" and "superintelligence" describe aspirational goals and do not indicate the presence of a real general intelligence. Use at your own risk. Nothing herein constitutes financial advice. MontrealAI and the maintainers accept no liability for losses incurred from using this software.

🏭 Production Deployment Guide — Alpha‑AGI Business v1

This guide summarises the minimal steps required to run the Alpha‑AGI Business v1 demo in a production‑like environment. The service works fully offline but upgrades automatically when OPENAI_API_KEY is present.

  1. Prepare the configuration
  2. Run python scripts/setup_config.py to create config.env if missing and edit as needed.
  3. Set OPENAI_API_KEY to enable cloud models or leave empty to use the bundled local fallbacks.
  4. Optionally enable the Google ADK gateway by setting ALPHA_FACTORY_ENABLE_ADK=true.
  5. Set MCP_ENDPOINT to push logs to a Model Context Protocol server (optional).
  6. Set MCP_TIMEOUT_SEC to adjust the timeout for MCP requests (default: 30 seconds).
  7. API_TOKEN defaults to "demo-token" (for demonstrations only); always set a strong secret before deploying.
  8. For API protection set either AUTH_BEARER_TOKEN or JWT_PUBLIC_KEY/JWT_ISSUER.
  9. Validate that all Python packages are available. From the project root run: bash AUTO_INSTALL_MISSING=1 python check_env.py --auto-install Provide WHEELHOUSE=/path/to/wheels for air‑gapped setups. Running this command is mandatory before executing the demos or running the test suite. The openai-agents and google-adk packages are optional and are only required when using the OpenAI Agents runtime or the Google ADK gateway.

  10. Build wheels for these optional packages when preparing an offline deployment: bash pip wheel openai-agents google-adk -w /path/to/wheels Provide this directory via WHEELHOUSE during installation on the production host. Run pre-commit run --all-files after the dependencies finish installing.

  11. Launch the service

  12. Docker (recommended for consistent environments): bash ./run_business_v1_demo.sh [--pull] [--gpu]

  13. Native Python: bash pip install -r ../../requirements.txt python run_business_v1_local.py --bridge

  14. Run in Colab

  15. Open colab_alpha_agi_business_v1_demo.ipynb.
  16. Run the setup cell; dependencies are installed automatically.

  17. The notebook exposes a Gradio dashboard and OpenAI Agents SDK bridge.

  18. Access the interface

  19. REST/Swagger docs: http://localhost:8000/docs
  20. Gradio dashboard: http://localhost:7860
  21. Prometheus metrics: http://localhost:8000/metrics

Verifying the ADK Gateway

When ALPHA_FACTORY_ENABLE_ADK=true and the optional google-adk package are installed, the service spawns an ADK gateway. Look for a log message like:

ADK gateway listening on http://0.0.0.0:${ALPHA_FACTORY_ADK_PORT}  (A2A protocol)

Confirm connectivity with:

curl http://localhost:${ALPHA_FACTORY_ADK_PORT}/docs
# or
curl http://localhost:${ALPHA_FACTORY_ADK_PORT}/healthz
  1. Shutting down
  2. Docker: ./run_business_v1_demo.sh --stop
  3. Native Python: press Ctrl+C; the orchestrator shuts down gracefully.

For advanced options see README.md in the same directory.

Further Resources