AI Service Onboarding¶
After building and testing your AI Service locally, follow these steps to onboard it into AIDF:
Onboarding Steps¶
- Role Switch — Assume the Platform Admin role in Infra Hub to manage service packaging and deployment
- Code Packaging — Organize code per AIDF folder structure → Create a ZIP archive of the entire service directory
- Publish to Artifactory — Use the Agile Workbench VS Code plugin to upload the ZIP to the Public Artifactory (Azure Blob container)
- Set Visibility — Choose Private (project-only) or Public (shared across workspace)
- Deploy via Manifest Engine — Click Deploy → Manifest Engine builds Docker image → pushes to ACR → deploys to AKS
- Test — Run a preliminary test with a sample payload to verify the service works correctly
- Admin Approval — A reviewer approves the service in My Reviews — only then is it available for pipeline use
Deployment Flow Detail¶
When you click Deploy in Infra Hub:
- Build Image: Converts the ZIP from Artifactory into a Docker image using the Dockerfile in
scripts/ - Push to ACR: Uploads the image to Azure Container Registry under a predictable naming scheme
- Deploy to AKS: Spins up the container on Azure Kubernetes Service, wiring in the correct HTTP route (
/datafactory/{component}/{serviceName}) from the manifest's component field