Build solution
Run the Full Pipeline
One command builds the solution including data processing and agent creation:
Fabric Workspace Mode
Note: If you omit
--fabric-workspace-id, the script will prompt you for it interactively. Press Enter key to start or Ctrl+C to cancel the process.
Azure Only Mode (if you ran azd env set AZURE_ENV_ONLY true before deploying)
Note: Press Enter key to start or Ctrl+C to cancel the process.
This uses the data/default folder and runs all setup steps:
| Step | What Happens | Time |
|---|---|---|
| 02 | Create Fabric Lakehouse, Ontology, and Data Agent | ~1.5min |
| 03 | Generate Agent Prompt | ~5s |
| 05 | Upload documents to AI Search | ~1min |
| 06 | Create Foundry Agent | ~10s |
Expected Output
> [02] Create Fabric Lakehouse & Load Data... OK (90.5s)
> [03] Generate Agent Prompt... OK (5.2s)
> [05] Upload to AI Search... OK (60.3s)
> [06] Create Foundry Agent... OK (10.1s)
> ✓ Done! 4/4 steps completed
> Next: python scripts/07_test_agent.py
This sets up entity types (Tickets, Inspections), data bindings from your Lakehouse tables, relationships between them, and the Fabric Data Agent.
Test the Fabric Data Agent
- Go to your Microsoft Fabric workspace.
- Open the Data Agent named
dataagent_<solution-name>_<suffix>. - If you do not see the Data Agent immediately, wait a few minutes and refresh the workspace.
- Ask test questions in the Data Agent or continue with the CLI-based agent test below.
Note: The Ontology setup may take up to 15 minutes to fully propagate, so retry after some time if you do not see good responses.
Test the Agent
Sample Conversation
============================================================
AI Agent Chat (Fabric SQL + Native Search)
============================================================
Chat Agent: dauypdob4c4d2k-ChatAgent
SQL Mode: Fabric Lakehouse
Lakehouse: <workshop_lakehouse_1>
Type 'quit' to exit, 'help' for sample questions
------------------------------------------------------------
You: What is the average score from inspections?
Agent: The average score from inspections is **77**.
You: What constitutes a failed inspection?
Agent: A failed inspection is defined by the following criteria:
1. **Score Requirement**: All inspections must achieve a minimum score of 80 out of 100.
2. **Follow-up Actions**: Inspections falling below this threshold will require immediate review and corrective action, along with follow-up assessments within 48 hours.
Teams must document all corrective actions taken.
You: Do any inspections violate quality control standards in our Inspection Procedures?
Agent: Yes, there are several inspections that violate the quality control standards as they scored below the required minimum of 80. Here are the details:
| Inspection ID | Customer Name | Result | Score |
|----------------|---------------|--------|-------|
| INSPECTION002 | Customer 4 | Pass | 65 |
| INSPECTION009 | Customer 10 | Pass | 60 |
| INSPECTION010 | Customer 5 | Pass | 71 |
| INSPECTION011 | Customer 6 | Pass | 74 |
| INSPECTION012 | Customer 1 | Fail | 69 |
| INSPECTION013 | Customer 15 | Pass | 69 |
| INSPECTION016 | Customer 10 | Pass | 64 |
| INSPECTION019 | Customer 11 | Pass | 73 |
| INSPECTION020 | Customer 6 | Pass | 72 |
| INSPECTION022 | Customer 2 | Pass | 62 |
| INSPECTION024 | Customer 7 | Fail | 70 |
| INSPECTION026 | Customer 3 | Pass | 67 |
| INSPECTION027 | Customer 5 | Pass | 69 |
| INSPECTION028 | Customer 8 | Pass | 62 |
| INSPECTION029 | Customer 10 | Pass | 75 |
| INSPECTION030 | Customer 11 | Pass | 60 |
| INSPECTION031 | Customer 8 | Pass | 60 |
| INSPECTION032 | Customer 3 | Pass | 62 |
| INSPECTION033 | Customer 2 | Pass | 68 |
| INSPECTION036 | Customer 15 | Pass | 62 |
| INSPECTION037 | Customer 3 | Pass | 67 |
| INSPECTION040 | Customer 4 | Pass | 74 |
These inspections need to be reviewed and may require corrective actions.
You: quit
Checkpoint
Solution Deployed!
You now have a working solution with:
- Data queries via Fabric IQ or Azure SQL function tools for AzureOnly mode
- Foundry IQ retrieving document knowledge
- Orchestrator Agent combining both sources