Nordlicht Reisen GmbH wants two environments, test and production, that arise from the same files, and new versions of the travel assistant (E-01) only after a green test set (R-10). This lesson designs the infrastructure and the path of an agent version up to its release. A Bicep draft and a small C# tool are compiled; the pipeline stays a sketch. You need an Azure account only for the optional run; costs then arise from the cost drivers: tokens and capacity of the deployments.

Prior knowledge. Bicep, what-if and environments are taught by "K3, lessons 3.2.2, 3.2.4 and 3.2.6"; managed identity "K3, lesson 3.3.1"; pipelines with GitHub Actions and Azure Pipelines "K3, lessons 3.7.1 and 3.7.2", approvals before the deploy "K3, lesson 3.7.6". Resource, project, deployment and role assignment as Bicep are shown by "AI track, lesson T10.1"; evaluation in the build "AI track, lesson T9.5". Whoever has them knows the basic framework. New here are: models as parameters per environment, the role of the project itself, the path of an agent version and what Learn leaves open about it.

The requirement

All numbers are assumptions of the task.

No. Requirement Type Measure Weight Source
1.2-01 Test and production arise from the same files. F 0 manual steps Must Operations
1.2-02 A model change is not a code change (R-08). Q 1 line in a parameter file Must Development
1.2-03 All resources are in location, preferably germanywestcentral (R-06). R 0 exceptions Must Data protection
1.2-04 A new agent version goes live only after a green test set. R 100 % of runs with a gate Must Quality
1.2-05 Rollback to the previous version. Q under 15 min Should Operations
1.2-06 The pipeline holds no secret (R-07). R 0 secrets Must Security