improve local testing and deployments with docker containers

Requirements:

• support Mac, Linux and Windows development environments
  • support VS Code IDE
• support Linux (Ubuntu) deployment to Digital Ocean, Google Cloud or LangSmith

Optional:

• on MacOS >= 26.0 (aka Tahoe) support Apple Containers (though without a compose solution)

It would be good to containerize the frontend and backend separately and manage them with a docker-compose.yml.

There should be a GitHub Action that builds x86 and arm64 docker images for deployment (triggered by merge to main or manual trigger on branch).

• The build should accept build time variables (e.g. TYPE=dev, TYPE=deploy) where
  • the dev TYPE installs dev tools for running tests, evaluations, etc.
  • the deploy TYPE should be a minimal image (strips out tests and utility scripts)
• Built images should be uploaded to ghcr.io
• The deployment should accept runtime variables and secrets (e.g. SHOW_MODEL_THINKING, MODEL_NAME, GOOGLE_APPLICATION_CREDENTIAL, LOG_LEVEL).

README.md, ARCHITECTURE.md & CLAUDE.md should be updated

Start by reading README.md, ARCHITECTURE.md, and CLAUDE.md to understand the current frontend, backend, testing, and deployment setup. Then plan the Dockerfiles, docker-compose.yml, and GitHub Action around the stated development, deployment, architecture, and runtime-variable requirements. Done means cross-platform development support, x86 and arm64 images uploaded to ghcr.io, deployment guidance, and updated documentation.