Skip to content

Challenge 4: Production Workflow

Time: ~20 minutes

Build a multi-agent orchestration workflow for TireForge Industries and take it to production.

Scenario

The individual agents you built in Challenge 1 are valuable — but in production, agents need to work together as an automated pipeline. In this challenge you wire the two agents into a full factory health workflow, run it from code, then build and test it visually in the Foundry portal.

deploy

Learning Objectives

  • Deploy persistent production agents (create once, reuse forever)
  • Orchestrate multiple agents step-by-step in a Python workflow
  • Build the same workflow visually in the Foundry portal
  • Invoke the portal workflow from Python with live streaming
  • View run history and traces in the portal

The Workflow

ensure_agents_deployed()
        |
        v
run_anomaly_scan()          <-- Anomaly Detection Agent checks all 5 machines
        |
        v (for each machine with anomalies)
run_fault_diagnosis()       <-- Fault Diagnosis Agent diagnoses root cause
        |
        v
print_factory_report()      <-- Consolidated Health Report

Part 1 — SDK: Build and Run the Python Workflow

Step 1: Review the implementation

Open deploy.py and review:

  • ensure_agents_deployed() — lists existing agents, creates anomaly-detection-agent and fault-diagnosis-agent if not present
  • run_anomaly_scan() — calls the anomaly agent, handles the check_thresholds function call loop
  • run_fault_diagnosis() — calls the diagnosis agent for each affected machine
  • run_factory_health_workflow() — orchestrates all steps and returns the consolidated report

Step 2: Run the workflow

cd factory/challenge-4-deploy
python deploy.py

Expected output: 

=== Step 1: Ensure Agents Are Deployed ===
  Found existing: anomaly-detection-agent
  Found existing: fault-diagnosis-agent

=== Step 2a: Anomaly Scan ===
  CP-003 CRITICAL: vibration 143% above max, pressure 13.8% above max, temperature 10.3% above max
  MX-001 WARNING: vibration 6.7% above max, temperature 2.6% above max
  IS-005 WARNING: vibration 30% above max
  ...

=== Step 2b: Fault Diagnosis ===
  Diagnosing CP-003...
  Diagnosing MX-001...
  Diagnosing IS-005...

TIREFORGE FACTORY HEALTH REPORT
  Machines checked   : 5
  Machines affected  : 2
  ...

Part 2 — Portal: Build and Test the Visual Workflow

Step 3: Verify agents are deployed in the portal

  1. Open the Microsoft Foundry portal
  2. Select your tire-factory-project
  3. Left sidebar → Build → Agents
  4. Confirm both agents appear:
  5. anomaly-detection-agent
  6. fault-diagnosis-agent

Step 4: Test the Anomaly Detection Agent

  1. Click anomaly-detection-agent → Playground
  2. Send: 
    Check machines MX-001 and CP-003 for anomalies.
  3. Watch the check_thresholds tool call fire for each machine
  4. Also try: 
    Check all machines: MX-001, EX-002, CP-003, CU-004, IS-005. Report every sensor out of spec.

Step 5: Test the Fault Diagnosis Agent

  1. Left sidebar → Agents → fault-diagnosis-agent → Playground
  2. Send: ``` Machine CP-003 has the following anomalies:
  3. temperature: 198.5 celsius (max 180, 10.3% above max)
  4. pressure: 18.2 bar (max 16.0, 13.8% above max)
  5. vibration: 7.3 mm/s (max 3.0, 143% above max) Diagnose the fault and recommend maintenance actions. ```
  6. Expected structure: LIKELY CAUSE / MAINTENANCE ACTIONS / URGENCY

Step 6: Build the workflow in the portal designer

  1. Left sidebar → Build → Workflows → + New workflow
  2. In the visual designer:
  3. + Add step → Agent → anomaly-detection-agent
    • Input: Run a full factory health check on all machines: MX-001, EX-002, CP-003, CU-004, IS-005.
  4. + Add step → Agent → fault-diagnosis-agent
    • Wire the output of the first step as input to this step
  5. Name it factory-health-workflow
  6. Click Save then Deploy

Step 7: Test the workflow in the portal playground

Why you must include the sensor data in your message

The agents use a check_thresholds tool that reads from a local Python file. The portal playground cannot execute Python functions — if you send a generic prompt, the agent will try to call the tool and stall waiting for a result that never arrives. Paste the sensor readings directly into your message so the agents can work without needing the tool.

  1. Open factory-health-workflow → Playground
  2. Paste the following message (data is pre-embedded so no tool calls are needed):
All sensor readings for today are below — analyse them directly, do not call check_thresholds.

MX-001 (mixer) — status: warning
  temperature: 92.3°C  [normal 60–90]  ⚠️ ABOVE MAX
  pressure:     3.1 bar [normal 2.0–4.0]
  vibration:    4.8 mm/s [normal 0–4.5]  ⚠️ ABOVE MAX
  rpm:          58 rpm  [normal 40–65]

EX-002 (extruder) — status: normal
  temperature: 115.0°C [normal 100–130]
  pressure:    12.5 bar [normal 10.0–15.0]
  vibration:    2.1 mm/s [normal 0–3.5]
  rpm:          30 rpm  [normal 20–40]

CP-003 (curing_press) — status: critical
  temperature: 198.5°C [normal 140–180]  🔴 ABOVE MAX
  pressure:    18.2 bar [normal 12.0–16.0]  🔴 ABOVE MAX
  vibration:    7.3 mm/s [normal 0–3.0]  🔴 ABOVE MAX
  rpm:           0 rpm  [normal 0]

CU-004 (cooling_unit) — status: normal
  temperature: 35.2°C  [normal 20–45]
  pressure:     1.0 bar [normal 0.8–1.5]
  vibration:    0.8 mm/s [normal 0–2.0]
  rpm:         120 rpm  [normal 80–150]

IS-005 (inspection_station) — status: warning
  temperature: 28.0°C  [normal 18–30]
  pressure:     1.0 bar [normal 0.8–1.2]
  vibration:    5.2 mm/s [normal 0–4.0]  ⚠️ ABOVE MAX
  rpm:        1800 rpm  [normal 1500–2200]

Detect all anomalies, then diagnose root causes and recommend remediation for affected machines.
  1. Watch the steps execute in sequence — anomaly scan first, then fault diagnosis
  2. Review the final consolidated report

Step 8: Invoke the portal workflow from Python (streaming)

Add to your .env: 

WORKFLOW_AGENT_NAME=factory-health-workflow

Re-run the script — Part B activates automatically: 

python deploy.py

You will see each workflow step appear live in the terminal: 

=== Invoking Portal Workflow: factory-health-workflow ===
  --> Step: anomaly_scan
      [completed] anomaly_scan
  --> Step: fault_diagnosis
      [completed] fault_diagnosis
<final report streamed here>

Step 9: View run history and traces

  1. Portal → your workflow → Run history tab
  2. Click the latest run to see the execution timeline — each step, duration, and output
  3. Left sidebar → Operate → Tracing to see the full distributed trace across both agent conversations

Success Criteria

  • [ ] Python workflow runs end-to-end: anomaly scan → diagnosis → factory health report
  • [ ] Both agents visible in the Foundry portal as persistent assets
  • [ ] Visual workflow created in the portal and tested in its playground
  • [ ] Portal workflow invoked from Python with live step streaming

Beyond the Lab: Production Deployment Options

You've built and tested your agents locally. Here's how to take them to production:

Option 1: Hosted Agents (What You Already Have)

Your agents created with agents.create_version() are already production-ready hosted agents. They live in Foundry indefinitely — any client can invoke them by name via the Responses API. No infrastructure to manage; Foundry handles scaling, versioning, and availability.

  • Versioning: Each create_version() produces an immutable version. Roll back by referencing an older version.
  • Multi-tenant: Multiple users/apps can call the same agent simultaneously.
  • Portal visibility: Agents appear under Build → Agents with playground, run history, and tracing.

Option 2: Foundry Workflows (Visual Orchestration)

What you built in Part 2 — wire multiple hosted agents into a DAG using the portal designer. The workflow becomes a deployable agent invoked via the same Responses API.

  • Step sequencing with automatic output passing
  • Streaming workflow_action events showing progress
  • Run history with per-step timing

Option 3: Azure App Service / Container Apps

Wrap your Python workflow in a FastAPI/Flask app for custom middleware, auth, or business logic:

# Example: FastAPI endpoint that calls your Foundry agents
@app.post("/factory-health-check")
async def health_check():
    report = run_factory_health_workflow(anomaly_agent, diagnosis_agent)
    return report

Deploy to App Service (managed PaaS) or Container Apps (auto-scaling containers).

Option 4: Azure Functions (Event-Driven)

Trigger agent workflows from events:

  • Timer trigger: Run the factory health check every hour
  • Service Bus trigger: Process each anomaly alert as it arrives from IoT Hub
  • HTTP trigger: On-demand endpoint for maintenance teams

Pay-per-execution, scales to zero when idle.

Option 5: CI/CD Quality Gates

Integrate evaluation into your deployment pipeline:

  • Run evaluate.py on every PR — block merge if quality drops below threshold
  • Promote agent versions: v1-dev → v1-staging → v1-prod after evaluation passes
  • Blue/green: Deploy new version to 10% traffic, compare metrics, then promote

Summary

Pattern Best For
Hosted Agents Always-on, invoke by name, no infra management
Foundry Workflows Multi-agent orchestration without code
App Service / Containers Custom auth, middleware, webhooks
Azure Functions Event-driven, pay-per-use, IoT integration
CI/CD Gates Automated quality assurance before promotion