This demo shows two things: how I use blue/green deployment in Azure Container Apps (ACA), and how I use feature flags in Azure App Configuration to do canary-style feature rollout.
The blue release is the stable production path and receives 100% of the site traffic, while the green release is used for testing and gradual feature exposure.
The idea behind blue/green is simple: keep two versions of the application available, one active and one ready to receive traffic when needed. That makes it easier to validate changes, switch versions safely, and roll back if something goes wrong. In my setup, the blue release is connected to my live site, while the green release is tested separately through a REST client.
On top of that, I use feature flags to control which AI features are available to which users. Instead of releasing everything to everyone at once, I can enable features selectively in the API based on the user. In this demo, sentiment analysis is available to everyone, language detection is available only to patrick, and summarization is available only to demo.
When a request comes in, the API checks the user and decides whether the feature is enabled. If the detect-language path is used, the API forwards the request to the llm-adapter, which then calls my Cloudflare Ollama Worker.
The Worker uses the /api/generate endpoint together with the prompt, and that prompt determines what the LLM will actually do.
This setup shows the difference between deployment strategy and feature rollout. Blue/green is about version control and safe rollback. Feature flags are about exposing functionality gradually and selectively. Together, they make it possible to deploy safely, test new behavior, keep production stable or when needed roll back fast.