Home / Resources / Resources — Reference Implementation Toolbox
Resources

Resources — Reference Implementation Toolbox

This directory is your toolbox for the RVAS. It contains a complete, working reference implementation of the AgentOps Control Tower — infrastructure-as-code, application code, Fabric setup automation, PySpark notebooks, pipelines, and a semantic model.

🧰 How to use it: Teams are encouraged to deploy the Azure foundation (agent workload + landing zone) from here so they can spend their creative energy on the Fabric Control Tower. Coaches hold the keys to the reference solution for each challenge and reveal it as needed — see the Coach Handbook.

What's here

Component Directory Used in What it provides
Agent Workload agent-workload/ Challenge 1 Full-stack Azure AI Foundry agent app — frontend (Next.js), backend & agent (FastAPI), behind API Management, backed by Cosmos DB, instrumented with Application Insights. Bicep IaC + azd manifest.
Observability Ingestion observability-ingestion/ Challenge 2 The ADLS Gen2 landing zone — Cost Management FOCUS exports, Resource Graph metadata, Log Analytics data export, and diagnostic settings. Bicep modules + Python export/validation scripts.
Fabric Control Tower fabric-control-tower/ Challenges 3–6 The Fabric layer — workspace/Lakehouse setup, OneLake shortcuts, Cosmos DB Mirroring, the Bronze/Silver/Gold notebooks, orchestration pipelines, and the Direct Lake semantic model.
Observability SDK observability-sdk/ Challenge 1 (reference) Shared Python library defining the agent telemetry contract — event schema, metrics, Azure Monitor wiring, and alert thresholds.

How it maps to the challenges

Challenge 1  ──▶  agent-workload/          (the telemetry SOURCE)
Challenge 2  ──▶  observability-ingestion/ (the landing ZONE)
Challenge 3  ──▶  fabric-control-tower/src/setup/   (shortcuts + Mirroring)
Challenge 4  ──▶  fabric-control-tower/fabric/      (medallion notebooks + pipelines)
Challenge 5  ──▶  fabric-control-tower/src/setup/semantic_model.json  (Direct Lake model)
Challenge 6  ──▶  fabric-control-tower/  +  Power BI / Data Activator  (operationalize)

Important notes

  • This is reference material, not a turnkey product. Names, regions, and IDs must be set for your environment. Read each component's README before deploying.
  • Deploy in order where there are dependencies: the agent workload and landing zone are independent of each other, but the Fabric Control Tower consumes both.
  • Cost & cleanup: everything is sized for a low-cost RVAS. Tear down with azd down --purge in each workload directory when you're finished (see prerequisites).
  • Same tenant: keep the Azure subscription and Fabric capacity in the same Entra tenant to avoid managed-identity and Mirroring friction.

Architecture

For how these components fit together end-to-end, see docs/architecture.md.

← Back to all content View source on GitHub ↗