docs(architect): weekly KB update — 52 files refreshed (2026-04)

Key content changes: - MLOps: MLflow 3 scorers expanded (RetrievalRelevance, Fluency, multi-turn judges) - MLflow 3 A/B eval: mirror_traffic GA confirmed, new scorer catalog - CI/CD: OIDC auth replaces deprecated --sdk-auth (Azure ML GitHub Actions) - Agent framework A2A: updated SDK patterns (A2ACardResolver, BearerAuth) - AG-UI backend tool rendering: accurate TOOL_CALL_* event shapes - Computer Use agents: US region requirement, credentials patterns - Purview governance: bulk term edit, expire/delete workflows - CAF AI Secure: 3-phase structure confirmed current - Copilot Studio: Claude Sonnet 4.5/4.6 GA, new orchestration controls - M365 manifest: v1.26 GA (April 2026), copilotAgents node - Power Platform: agent flow capacity enforcement corrected - Azure Monitor: Simple Log Alerts GA, AMBA for policy-based alerting - Security Copilot: SCU capacity model (400 SCU/1000 users) - EU Data Boundary: all EU + EFTA countries confirmed - gateway-multi-backend: added 4th topology, subscription-level quota note Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-10 11:31:11 +02:00 · 2026-04-10 11:31:11 +02:00 · be4925a8ff
commit be4925a8ff
parent 6645e93205
40 changed files with 398 additions and 239 deletions
--- a/skills/ms-ai-engineering/references/mlops-genaiops/azure-ml-pipelines-orchestration.md
+++ b/skills/ms-ai-engineering/references/mlops-genaiops/azure-ml-pipelines-orchestration.md
@ -1,6 +1,6 @@
 # Azure ML Pipelines - Orchestration and Automation

-**Last updated:** 2026-02
+**Last updated:** 2026-04
 **Verified:** MCP 2026-04
 **Status:** GA
 **Category:** MLOps & GenAIOps
@ -22,59 +22,81 @@ Fra et kostnads- og effektivitetsperspektiv gir pipelines betydelige fordeler: d
 ### Pipeline Components (v2)


-### Azure ML Pipelines — Python SDK v2 (Tutorial 2026)
+### Azure ML Pipelines — Python SDK v2 (Tutorial, Verified MCP 2026-04)

-**Key benefits**: Standardized MLOps, scalable team collaboration, training efficiency, cost reduction.
+**Key benefits**: Standardized MLOps practice, scalable team collaboration, training efficiency, cost reduction.

-**Pipeline creation pattern** (SDK v2):
+**Pipeline creation pattern** (SDK v2 — from official tutorial):
 ```python
 from azure.ai.ml import MLClient, dsl, Input, Output, command
-from azure.identity import DefaultAzureCredential
+from azure.identity import DefaultAzureCredential, InteractiveBrowserCredential

-ml_client = MLClient(DefaultAzureCredential(), subscription_id, resource_group, workspace)
+try:
+    credential = DefaultAzureCredential()
+    credential.get_token("https://management.azure.com/.default")
+except Exception:
+    credential = InteractiveBrowserCredential()

-# 1. Create reusable components
+ml_client = MLClient(credential, subscription_id, resource_group, workspace)
+# Note: MLClient initialization is lazy — no connection until first call
+
+# 1. Create reusable components (programmatic definition)
 data_prep_component = command(
-    name="data_prep",
+    name="data_prep_credit_defaults",
    inputs={"data": Input(type="uri_folder"), "test_train_ratio": Input(type="number")},
-    outputs={"train_data": Output(type="uri_folder"), "test_data": Output(type="uri_folder")},
+    outputs={"train_data": Output(type="uri_folder", mode="rw_mount"),
+             "test_data": Output(type="uri_folder", mode="rw_mount")},
    code="./components/data_prep",
-    command="python data_prep.py --data ${{inputs.data}} ...",
-    environment=f"{env.name}:{env.version}",
+    command="python data_prep.py --data ${{inputs.data}} --test_train_ratio ${{inputs.test_train_ratio}} ...",
+    environment=f"{pipeline_job_env.name}:{pipeline_job_env.version}",
 )
 # Register for reuse
-ml_client.create_or_update(data_prep_component.component)
+data_prep_component = ml_client.create_or_update(data_prep_component.component)

 # 2. Define pipeline with @dsl.pipeline decorator
-@dsl.pipeline(compute="serverless", description="E2E training pipeline")
-def training_pipeline(data_input, test_train_ratio, learning_rate, model_name):
+@dsl.pipeline(
+    compute="serverless",  # "serverless" runs on serverless compute
+    description="E2E data_prep-train pipeline",
+)
+def credit_defaults_pipeline(data_input, test_train_ratio, learning_rate, registered_model_name):
    prep_job = data_prep_component(data=data_input, test_train_ratio=test_train_ratio)
    train_job = train_component(
        train_data=prep_job.outputs.train_data,
        test_data=prep_job.outputs.test_data,
        learning_rate=learning_rate,
-        registered_model_name=model_name,
+        registered_model_name=registered_model_name,
    )
+    return {
+        "pipeline_job_train_data": prep_job.outputs.train_data,
+        "pipeline_job_test_data": prep_job.outputs.test_data,
+    }

 # 3. Submit pipeline
 pipeline_job = ml_client.jobs.create_or_update(
-    training_pipeline(data_input=..., ...),
-    experiment_name="e2e_pipeline"
+    credit_defaults_pipeline(
+        data_input=Input(type="uri_file", path=credit_data.path),
+        test_train_ratio=0.25,
+        learning_rate=0.05,
+        registered_model_name="credit_defaults_model",
+    ),
+    experiment_name="e2e_registered_components"
 )
+ml_client.jobs.stream(pipeline_job.name)
 ```

 **Component lifecycle**:
-1. Write YAML spec or create programmatically (`CommandComponent`)
-2. Register with name+version in workspace or registry
-3. Load and compose into pipeline
-4. Submit via `ml_client.jobs.create_or_update()`
+1. Write YAML spec (`train.yml`) or create programmatically (`CommandComponent` / `command()`)
+2. Register with name+version: `ml_client.create_or_update(component)`
+3. Load and compose into pipeline using `@dsl.pipeline` decorator
+4. Submit via `ml_client.jobs.create_or_update()` with experiment name

-**Compute options**: `serverless` (recommended), named compute cluster, or per-step compute override.
-**Environment**: Curated environments (`azureml://registries/azureml/environments/sklearn-1.5/labels/latest`) or custom conda/Docker.
+**Compute options**: `serverless` (recommended — zero config), named compute cluster, or per-step compute override (e.g., `train_step.compute = "cpu-cluster"`).
+**Environment**: Curated environments (`azureml://registries/azureml/environments/sklearn-1.0/labels/latest`) or custom conda/Docker (base image: `mcr.microsoft.com/azureml/openmpi4.1.0-ubuntu22.04:latest`).
 **Output types**: `uri_folder` (data), `mlflow_model` (model), `uri_file` (file).

-**MLflow integration**: Use `mlflow.start_run()` in scripts for automatic experiment tracking (metrics, parameters, models).
+**MLflow integration**: Use `mlflow.start_run()` + `mlflow.sklearn.autolog()` in training scripts for automatic experiment tracking. Models registered via `mlflow.sklearn.log_model()` with `registered_model_name`.

+**VNet note**: If workspace uses a managed virtual network, add outbound rules to allow access to public Python package repositories.

 | Komponent-type | Beskrivelse | Bruksområde |
 |----------------|-------------|-------------|
@ -564,41 +586,41 @@ Er det >3 steg i workflow?

 1. **What are Azure Machine Learning pipelines?**
   https://learn.microsoft.com/en-us/azure/machine-learning/concept-ml-pipelines?view=azureml-api-2
-   *Confidence: Verified (Feb 2026)*
+   *Confidence: Verified (April 2026)*

 2. **Schedule machine learning pipeline jobs**
   https://learn.microsoft.com/en-us/azure/machine-learning/how-to-schedule-pipeline-job?view=azureml-api-2
-   *Confidence: Verified (Feb 2026)*
+   *Confidence: Verified (April 2026)*

 3. **Create and run machine learning pipelines using components with the Azure Machine Learning SDK v2**
   https://learn.microsoft.com/en-us/azure/machine-learning/how-to-create-component-pipeline-python?view=azureml-api-2
-   *Confidence: Verified (Feb 2026)*
+   *Confidence: Verified (April 2026)*

 4. **Tutorial: Create production machine learning pipelines**
   https://learn.microsoft.com/en-us/azure/machine-learning/tutorial-pipeline-python-sdk?view=azureml-api-2
-   *Confidence: Verified (Feb 2026)*
+   *Confidence: Verified (April 2026)*

 5. **Use parallel jobs in pipelines**
   https://learn.microsoft.com/en-us/azure/machine-learning/how-to-use-parallel-job-in-pipeline?view=azureml-api-2
-   *Confidence: Verified (Feb 2026)*
+   *Confidence: Verified (April 2026)*

 6. **Manage inputs and outputs for components and pipelines**
   https://learn.microsoft.com/en-us/azure/machine-learning/how-to-manage-inputs-outputs-pipeline?view=azureml-api-2
-   *Confidence: Verified (Feb 2026)*
+   *Confidence: Verified (April 2026)*

 7. **Create jobs and input data for batch endpoints**
   https://learn.microsoft.com/en-us/azure/machine-learning/how-to-access-data-batch-endpoints-jobs?view=azureml-api-2
-   *Confidence: Verified (Feb 2026)*
+   *Confidence: Verified (April 2026)*

 8. **Upgrade pipeline endpoints to SDK v2**
   https://learn.microsoft.com/en-us/azure/machine-learning/migrate-to-v2-deploy-pipelines?view=azureml-api-2
-   *Confidence: Verified (Feb 2026)*
+   *Confidence: Verified (April 2026)*

 ### Code Samples (Verified)

 - **Azure ML Examples Repository (azureml-examples/sdk/python/schedules):**
  https://github.com/Azure/azureml-examples
-  *Confidence: Verified (Feb 2026)*
+  *Confidence: Verified (April 2026)*

 ### Konfidensgradering per seksjon

@ -613,5 +635,5 @@ Er det >3 steg i workflow?
 | Kostnad og lisensiering | Verified + Baseline | MS Learn: cost considerations + Azure pricing |
 | For arkitekten | Baseline | Arkitekturkonsulent-erfaring |

-**Verified:** Informasjon hentet direkte fra Microsoft Learn MCP-dokumentasjon (februar 2026).
+**Verified:** Informasjon hentet direkte fra Microsoft Learn MCP-dokumentasjon (april 2026).
 **Baseline:** Informasjon basert på modellkunnskap og arkitekturerfaring, konsistent med Azure ML prinsipper.