microsoft/frontier-fabric-agentops-rvas

microsoft/frontier-fabric-agentops-rvas

Description: AgentOps Control Tower (Fabric Edition)

Language: HTML

License: MIT

Stars: 0

Forks: 2

Open issues: 3

Created: 2026-06-01T10:16:28Z

Pushed: 2026-09-12T07:59:55Z

Default branch: main

Fork: no

Archived: no

README:

Frontier Fabric AgentOps RVAS

Build an AgentOps Control Tower on Microsoft Fabric

> A hands-on, challenge-based RVAS (Real Value Acceleration Solutions) where your team builds a Fabric-based control tower that > ingests and correlates telemetry from Azure services and Foundry-built agents — enabling > end-to-end monitoring, analytics, and dashboards for reliability, cost, and performance.

> 🌐 Explore it online: browse every challenge, coach guide, doc, and reference file — without > leaving the browser — at **. Running > this for a customer? Use the delivery composer** > to pick the challenges that fit their environment and export a tailored delivery brief.

---

The Mission

Your customer has gone all-in on AI agents. Teams across the business are shipping agents built on Azure AI Foundry, running on Azure Container Apps, and backed by Cosmos DB. Adoption is exploding — and so is the blind spot.

Nobody can answer the questions leadership keeps asking:

• Reliability — Which agents are healthy? Where are errors and latency spikes coming from?
• Cost — What is each agent costing us? Which team or use case owns that spend?
• Performance — Are we meeting SLAs? How many tokens are we burning, and is it trending up?

The telemetry exists, but it is scattered across Application Insights, Azure Monitor, Cost Management, Resource Graph, and Cosmos DB. The customer has chosen Microsoft Fabric to unify it all into a single AgentOps Control Tower.

That is your job. Over the course of this RVAS your team will stand up the end-to-end platform: light up a live agent workload, land its telemetry in a lake, connect it to Fabric with zero-copy OneLake shortcuts and Cosmos DB Mirroring, refine it through a medallion architecture, and deliver the Control Tower dashboards that finally give the business a single pane of glass.

This is not a guided click-through. Each challenge gives you a goal and success criteria — *how* you get there is up to your team. Coaches are in the room to unblock you, not to hand you the answer.

---

What You'll Build

graph TB
subgraph SRC["Telemetry Sources (Azure)"]
FOUNDRY["Azure AI Foundry
(agent models)"]
ACA["Container Apps
FE · BE · Agent"]
CDB["Cosmos DB
(conversations)"]
APPI["Application Insights"]
ACM["Cost Management
(FOCUS export)"]
ARG["Resource Graph
(metadata)"]
ACA FOUNDRY
ACA --> CDB
ACA --> APPI
end
subgraph LZ["Landing Zone"]
ADLS["ADLS Gen2
costs · metrics · logs · metadata · diagnostics"]
end
subgraph FAB["Microsoft Fabric — AgentOps Control Tower"]
subgraph ONE["OneLake"]
SC["Shortcuts (zero-copy)"]
MIR["Cosmos DB Mirroring"]
end
subgraph MED["Medallion Lakehouse"]
BRZ["Bronze (raw)"] --> SLV["Silver (curated)"] --> GLD["Gold (data products)"]
end
SEM["Direct Lake
Semantic Model"]
SC --> BRZ
MIR --> BRZ
GLD --> SEM
end
subgraph SERVE["Control Tower"]
PBI["Power BI Dashboards
Reliability · Cost · Performance"]
ACT["Real-time Alerts
(Data Activator)"]
end
APPI --> ADLS
ACM --> ADLS
ARG --> ADLS
ADLS --> SC
CDB -. mirror .-> MIR
SEM --> PBI
GLD --> ACT

By the end you will have a working pipeline that flows from a user's click, through agent inference and tool calls, into OneLake, and out to executive dashboards and live alerts — all on first-party Microsoft technology.

---

RVAS Format

| | | |---|---| | Style | Team-based (3–5 people), coach-facilitated, challenge-driven | | Level | 200–400 (intermediate to advanced) | | Duration | 1.5–2 days (7 challenges, ~1.5–2.5 hrs each; Challenge 6 is a stretch) | | Environment | The customer's own Azure subscription(s) + a Microsoft Fabric capacity (trial capacity is fine) | | Roles | Platform/Cloud engineer · Data engineer · FinOps practitioner · AI engineer (mix per team) |

Teams progress through the challenges in order — each one builds on the last. You are encouraged to divide and conquer within a challenge, then regroup at each success checkpoint.

---

The Challenges

| # | Challenge | What you'll achieve | Primary skills | |---|---|---|---| | 0 | [Mission Briefing & Environment Setup](challenges/challenge-00-mission-setup.md) | Azure + Fabric trial capacity ready, tooling installed, team onboarded | Setup, Fabric admin | | 1 | [Light Up the Agents](challenges/challenge-01-agent-telemetry.md) | Deploy the Foundry agent workload and confirm telemetry is flowing | Azure, Container Apps, App Insights | | 2 | [Build the Telemetry Landing Zone](challenges/challenge-02-landing-zone.md) | Land cost, metrics, logs, metadata & diagnostics in ADLS Gen2 | Azure Monitor, FinOps, IaC | | 3 | [Connect Fabric to the Enterprise](challenges/challenge-03-onelake-foundation.md) | Lakehouse + OneLake shortcuts + Cosmos DB Mirroring (zero-copy) | Fabric, OneLake | | 4 | [Refine the Signal](challenges/challenge-04-medallion-pipeline.md) | Bronze→Silver→Gold medallion pipeline correlating all sources | PySpark, data engineering | | 5 | [Stand Up the Control Tower](challenges/challenge-05-control-tower-dashboards.md) | Direct Lake semantic model + Reliability/Cost/Performance dashboards | Power BI, DAX, semantic modeling | | 6 | [Make It Operational](challenges/challenge-06-operationalize.md) 🏆 | Stretch: real-time alerts, FinOps chargeback, RLS, automation | Data Activator, FinOps, governance |

See the [Challenge Index](challenges/README.md) for the full arc, dependencies, and scoring.

---

Getting Started

1. Read the mission (above) and the [reference architecture](docs/architecture.md). 2. Confirm prerequisites — work through [docs/prerequisites.md](docs/prerequisites.md) with your coach before day 1. This covers Azure roles, Fabric trial capacity, and local tooling. 3. Start with [Challenge 0](challenges/challenge-00-mission-setup.md). 4. Keep the [FAQ](docs/faq.md) handy.

> 💡 Coaches: start at the [Coach Handbook](coach/README.md). Each challenge has a matching coach > guide with the reference solution, key commands, common pitfalls, and talking points.

---

##...