Quota warning suggests wrong region variable for AI model deployment

[agxc]

Describe the bug

During azd up, the model-quota preflight correctly reports that gpt-5.4-mini (GlobalStandard) has no available quota in australiaeast, but its remediation suggests:

azd env set AZURE_LOCATION <location>

This repository separates the non-AI resource region (AZURE_LOCATION) from the Foundry/model region (AZURE_AI_SERVICE_LOCATION). Following the warning would move Container Apps, Functions, Storage, and database resources instead of changing only the AI deployment region.

Expected behavior

The warning should suggest:

azd env set AZURE_AI_SERVICE_LOCATION <location-with-quota>

Alternatively, the message should distinguish the AI model location from the general infrastructure location.

How does this bug make you feel?

The remediation is misleading because it changes the deployment topology for unrelated resources.

---

Debugging information

Steps to reproduce

1. Use the current main branch.
2. Select australiaeast for both AZURE_LOCATION and AZURE_AI_SERVICE_LOCATION.
3. Set AZURE_GPT_MODEL_NAME to gpt-5.4-mini and AZURE_GPT_MODEL_VERSION to 2026-03-17.
4. Use the default GlobalStandard SKU and 50K TPM capacity.
5. Run azd up with a subscription that has zero available quota for that model in Australia East.
6. Observe the suggested AZURE_LOCATION command.

Environment

• OS: Linux dev container
• azd: 1.29.0
• Model: gpt-5.4-mini
• Version: 2026-03-17
• SKU: GlobalStandard
• AI region: australiaeast
• Requested capacity: 50K TPM
• Available capacity: 0

Logs

(!) Warning: Insufficient quota for model "gpt-5.4-mini" (SKU: GlobalStandard) in australiaeast
Requested: 50 · Available: 0
Suggestion: No quota is available. Change your deployment location via azd env set AZURE_LOCATION <location> or request a quota increase in the Azure portal.

Tasks

To be filled in by the engineer picking up the issue

• Update the quota-warning remediation to reference the AI-specific location variable.
• Add coverage for templates with separate infrastructure and AI regions.