feat: dynamic parameter references in usageName metadata for quota-based location filtering by Copilot · Pull Request #6928 · Azure/azure-dev

Copilot · 2026-02-27T19:16:46Z

Original prompt

Take the changes to create a PR that fixes #5432

In the issue, there is a linked azd template that uses static quota checks (https://github.com/microsoft/Multi-Agent-Custom-Automation-Engine-Solution-Accelerator/blob/main/infra/main.bicep#L39). Mentioned how this sample can be updated to use:
@metadata({
  azd: {
    type: 'location'
    usageName: [
      '$(p:gptModelName), $(p:gptModelCapacity)'
      '$(p:gpt4_1ModelName), $(p:gpt4_1ModelCapacity)'
      '$(p:gptReasoningModelName), $(p:gptReasoningModelCapacity)'
    ]
  }
})
And that would make it dynamic quota checks that depends on the parameters defined for models.

The user has attached the following file paths as relevant context:

cli/azd/AGENTS.md

Created from VS Code.

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

… quota checks (#5432) Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

… unnecessary code Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

…sed location filtering Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

vhvb1989 · 2026-02-27T19:36:26Z

cli/azd/extensions/azure.appservice/go.mod

 go 1.25

 require (
-	github.com/Azure/azure-sdk-for-go/sdk/azcore v1.20.0


@copilot remove all the changes to extensions folder /extensions
This PR should not be updating anything from the extensions folder

Done — reverted the unrelated go.mod/go.sum changes to extensions/azure.appservice and extensions/microsoft.azd.concurx in commit e219a7c.

Copilot

Pull request overview

This PR enables dynamic quota-based location filtering in azd templates by allowing usageName metadata to reference other template parameters (e.g., model name and capacity), then resolving those into SKU usage names via the AI model catalog before performing quota checks.

Changes:

Added $(p:paramName) reference parsing/resolution utilities and metadata dependency extraction.
Introduced dependency-aware parameter prompt ordering (topological sort) so referenced parameters are prompted before dependents.
Implemented SKU usage-name auto-resolution for referenced usageName entries, plus supporting docs/tests and changelog entry.

Reviewed changes

Copilot reviewed 9 out of 11 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
cli/azd/pkg/infra/provisioning/bicep/prompt.go	Resolves `$(p:...)` in `usageName`, auto-resolves model SKU usage names, and wires resolved usage names into quota-based location filtering.
cli/azd/pkg/infra/provisioning/bicep/bicep_provider.go	Adds topo-sorted prompting and a `resolvedValues` map to support dependency-aware prompting and reference resolution.
cli/azd/pkg/infra/provisioning/bicep/prompt_test.go	Updates prompt tests for signature changes and adds tests for topo-sort ordering/error cases.
cli/azd/pkg/azure/arm_template.go	Adds `HasParamReferences`, `ExtractParamReferences`, `ResolveParamReferences`, and `AzdMetadata.ParamDependencies()`.
cli/azd/pkg/azure/arm_template_test.go	Adds unit tests for parameter reference detection, extraction, resolution, and dependency collection.
cli/azd/docs/design/metadata-parameter-references.md	Design doc describing reference syntax, SKU resolution behavior, ordering, and intended UX/errors.
cli/azd/CHANGELOG.md	Documents the new `usageName` parameter reference feature for location parameters.
cli/azd/extensions/microsoft.azd.concurx/go.mod	Updates extension dependencies (large tidy/bump set) not directly tied to this feature.
cli/azd/extensions/microsoft.azd.concurx/go.sum	Corresponding dependency checksum updates for concurx extension.
cli/azd/extensions/azure.appservice/go.mod	Updates extension dependencies (large tidy/bump set) not directly tied to this feature.
cli/azd/extensions/azure.appservice/go.sum	Corresponding dependency checksum updates for azure.appservice extension.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

cli/azd/pkg/infra/provisioning/bicep/prompt.go

Copilot · 2026-02-27T19:38:27Z

cli/azd/pkg/infra/provisioning/bicep/prompt.go

+		// Parse the resolved string as "modelName, capacity" or "modelName"
+		parts := strings.SplitN(resolved, ",", 2)
+		modelName := strings.TrimSpace(parts[0])
+		capacityStr := ""
+		if len(parts) == 2 {
+			capacityStr = strings.TrimSpace(parts[1])
+		}
+
+		// Resolve the model name to its full SKU usage name
+		skuUsageName, err := p.resolveModelSkuUsageName(ctx, p.env.GetSubscriptionId(), modelName)
+		if err != nil {
+			return nil, fmt.Errorf("resolving SKU for model '%s': %w", modelName, err)
+		}


Any presence of $(p:...) triggers “model name → SKU usage name” catalog lookup, even if the template author only intended to reference capacity (e.g. OpenAI.GlobalStandard.gpt-4o, $(p:cap)). In that case, the code will treat the full usage name as a model name and likely fail the catalog lookup. Consider supporting mixed cases by detecting when the resolved first token already looks like a full usage name (or providing an explicit opt-in marker), and skipping SKU auto-resolution in that scenario.

@copilot apply changes based on this feedback

Applied in commit c9b8afc. After resolving $(p:...) references, the code now checks whether the first token already looks like a full SKU usage name by detecting a dot (Provider.SKU.ModelName format). If so, the AI catalog lookup is skipped and the token is used as-is — so OpenAI.GlobalStandard.gpt-4o, $(p:cap) works correctly. Two new unit tests cover this path.

Copilot · 2026-02-27T19:38:27Z

cli/azd/pkg/infra/provisioning/bicep/prompt.go

+// resolveModelSkuUsageName resolves a model name to its full SKU usage name by querying the
+// AI model catalog. It filters SKUs to only standard-tier types (GlobalStandard, DataZoneStandard,
+// Standard), excluding provisioned, batch, and developer SKUs which have fundamentally different
+// billing and usage models. If GlobalStandard is available, it is selected automatically as it
+// offers the highest quota and broadest availability. Otherwise, the user is prompted to choose
+// among the remaining standard-tier SKUs.
+func (p *BicepProvider) resolveModelSkuUsageName(
+	ctx context.Context,
+	subId string,
+	modelName string,
+) (string, error) {
+	if p.aiModelService == nil {
+		return "", fmt.Errorf("AI model service is not configured")
+	}
+
+	models, err := p.aiModelService.ListModels(ctx, subId, nil)
+	if err != nil {
+		return "", fmt.Errorf("listing AI models: %w", err)
+	}
+
+	// Find the target model and collect unique SKUs across all versions
+	type skuInfo struct {
+		Name      string
+		UsageName string
+	}
+	var allSkus []skuInfo
+	seen := map[string]struct{}{}
+	modelFound := false
+
+	for _, model := range models {
+		if model.Name != modelName {
+			continue
+		}
+		modelFound = true
+		for _, version := range model.Versions {
+			for _, sku := range version.Skus {
+				if _, ok := seen[sku.UsageName]; !ok {
+					seen[sku.UsageName] = struct{}{}
+					allSkus = append(allSkus, skuInfo{
+						Name:      sku.Name,
+						UsageName: sku.UsageName,
+					})
+				}
+			}
+		}
+		break
+	}
+
+	if !modelFound {
+		return "", fmt.Errorf("model '%s' not found in the AI model catalog", modelName)
+	}
+	if len(allSkus) == 0 {
+		return "", fmt.Errorf("no SKUs found for model '%s'", modelName)
+	}
+
+	// Filter to standard-tier SKUs only.
+	// Provisioned (PTU-based), Batch (async 24-hr), and Developer SKUs have fundamentally
+	// different billing and usage models that are not suitable for typical deployments.
+	standardTierSkus := make([]skuInfo, 0, len(allSkus))
+	for _, s := range allSkus {
+		switch s.Name {
+		case "GlobalStandard", "DataZoneStandard", "Standard":
+			standardTierSkus = append(standardTierSkus, s)
+		}
+	}
+
+	// If no standard-tier SKUs are available, fall back to all SKUs so the user
+	// can still proceed (they may have a provisioned deployment in mind).
+	candidates := standardTierSkus
+	if len(candidates) == 0 {
+		candidates = allSkus
+	}
+
+	if len(candidates) == 1 {
+		return candidates[0].UsageName, nil
+	}
+
+	// Auto-select GlobalStandard if available — it has the highest default quota
+	// and broadest availability, recommended as the starting point for most workloads.
+	for _, s := range candidates {
+		if s.Name == "GlobalStandard" {
+			return s.UsageName, nil
+		}
+	}
+
+	// Multiple candidates without GlobalStandard — prompt the user to select one
+	options := make([]string, len(candidates))
+	for i, s := range candidates {
+		options[i] = fmt.Sprintf("%s (%s)", s.Name, s.UsageName)
+	}
+
+	choice, err := p.console.Select(ctx, input.ConsoleOptions{
+		Message: fmt.Sprintf("Select a deployment SKU for model '%s':", modelName),
+		Options: options,
+	})
+	if err != nil {
+		return "", err
+	}
+
+	return candidates[choice].UsageName, nil
+}
+
+// resolveUsageNamesWithReferences resolves usageName entries that contain $(p:...) parameter
+// references. For referenced entries, the resolved value is treated as "modelName, capacity"
+// and the model name is resolved to its full SKU usage name via the AI model catalog.
+// Constant (non-referenced) entries are passed through unchanged.
+func (p *BicepProvider) resolveUsageNamesWithReferences(
+	ctx context.Context,
+	usageNames []string,
+	resolvedValues map[string]any,
+) ([]string, error) {


New behavior adds SKU auto-resolution (resolveModelSkuUsageName) and reference expansion (resolveUsageNamesWithReferences), but there are no unit tests covering these paths (e.g., multiple SKUs prompting, GlobalStandard auto-selection, model-not-found error, and end-to-end quota filtering using $(p:...)). Adding focused tests with a mocked aiModelService and console selection would help prevent regressions and validate the intended UX.

cli/azd/extensions/microsoft.azd.concurx/go.mod

cli/azd/extensions/azure.appservice/go.mod

Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

…full usage name Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

Checkpoint from VS Code for cloud agent session

1b4ac36

Copilot AI assigned Copilot and vhvb1989 Feb 27, 2026

Copilot started work on behalf of vhvb1989 February 27, 2026 19:16 View session

Copilot AI and others added 3 commits February 27, 2026 19:21

feat: dynamic parameter references in usageName metadata for location…

ac90bdd

… quota checks (#5432) Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

fix: address code review feedback - simplify error message and remove…

d33185c

… unnecessary code Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

feat: dynamic parameter references in usageName metadata for quota-ba…

ffff26d

…sed location filtering Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

Copilot AI changed the title ~~[WIP] Fix static quota checks to use dynamic quota checks~~ feat: dynamic parameter references in usageName metadata for quota-based location filtering Feb 27, 2026

Copilot finished work on behalf of vhvb1989 February 27, 2026 19:28

vhvb1989 marked this pull request as ready for review February 27, 2026 19:32

vhvb1989 requested a review from wbreza as a code owner February 27, 2026 19:32

Copilot AI review requested due to automatic review settings February 27, 2026 19:32

vhvb1989 requested review from JeffreyCA, hemarina, rajeshkamal5050, tg-msft, vhvb1989 and weikanglim as code owners February 27, 2026 19:32

Copilot started reviewing on behalf of vhvb1989 February 27, 2026 19:33 View session

vhvb1989 reviewed Feb 27, 2026

View reviewed changes

Copilot started work on behalf of vhvb1989 February 27, 2026 19:36 View session

Copilot AI reviewed Feb 27, 2026

View reviewed changes

revert: remove unrelated extension go.mod/go.sum changes

e219a7c

Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

Copilot finished work on behalf of vhvb1989 February 27, 2026 19:39

Copilot started work on behalf of vhvb1989 March 1, 2026 03:56 View session

Copilot AI and others added 2 commits March 1, 2026 04:01

fix: skip SKU catalog lookup when usageName first token is already a …

c9b8afc

…full usage name Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

Co-authored-by: vhvb1989 <24213737+vhvb1989@users.noreply.github.com>

9794b84

Copilot finished work on behalf of vhvb1989 March 1, 2026 04:05

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: dynamic parameter references in usageName metadata for quota-based location filtering#6928

feat: dynamic parameter references in usageName metadata for quota-based location filtering#6928
Copilot wants to merge 7 commits intomainfrom
copilot/vscode-mm59wsrj-ulwy

Copilot AI commented Feb 27, 2026 •

edited

Loading

Uh oh!

vhvb1989 Feb 27, 2026

Uh oh!

Copilot AI Feb 27, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Copilot AI Feb 27, 2026

Uh oh!

vhvb1989 Mar 1, 2026

Uh oh!

Copilot AI Mar 1, 2026

Uh oh!

Copilot AI Feb 27, 2026

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

Copilot AI commented Feb 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

vhvb1989 Feb 27, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 27, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Copilot AI Feb 27, 2026

Choose a reason for hiding this comment

Uh oh!

vhvb1989 Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 27, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Copilot AI commented Feb 27, 2026 •

edited

Loading