Azure/openai Usage

2026.08.21.1

Changed: Added .describe(...) documentation to previously undocumented fields in DeploymentUsageSchema, ResourceUsageSchema, ResourceListSchema, and ScanResultsSchema (deployment/resource identity fields, token counts, period and rate fields, and the truncated flag). Tightened clientSecret in GlobalArgsSchema to require a non-empty string — an empty secret can never succeed against Azure AD's token endpoint. No behavioral changes.

2026.08.14.4

Fixed: three issues surfaced after the #360 dedup fix. First, the ARM $filter=kind eq 'OpenAI' or kind eq 'AIServices' query param sent to the Cognitive Services list endpoint is not reliably enforced server-side, so non-token-emitting kinds (Face, ComputerVision, TextAnalytics, etc.) leaked into discovery and wasted metrics attempts; listAiResources now also filters client-side. Second, resources with zero usage in the lookback window were silently dropped with no log line at all, indistinguishable from a resource that was never attempted; scan_subscriptions now logs a debug line before each metrics attempt and an info line for zero-usage outcomes. Third, per-resource metrics failures logged only String(err), discarding detail needed to root-cause failures that don't reproduce via az rest; failure logs now include the resource's subscription/resourceGroup/kind and the raw error's name/message/stack.

2026.08.14.3

Fixed: listAiResources could enumerate the same ARM resource twice across paginated nextLink requests, causing scan_subscriptions to process more candidates than the deduplicated resource count reported by list_ai_resources — surfacing as some resources failing to scan and the result being marked truncated: true even though nothing was actually throttled or missing. Results are now deduped by ARM resource ID across pages. Per-page fetch counts and first/last resource IDs are now logged at debug level to make a repeat of this easier to diagnose from logs alone.

2026.08.14.1

Fixed: listAiResources ignored the ARM API's nextLink pagination cursor, silently dropping resources past the first page. GET .../Microsoft.CognitiveServices/accounts paginates at a small page size (observed: 2 per page); a subscription with 10 OpenAI/AIServices accounts was reporting only 1 in scan_subscriptions and list_ai_resources output. Both methods now follow nextLink until exhausted.

2026.07.31.1

Fixed: README incorrectly stated authentication uses az CLI (az login). The extension actually uses Azure AD client credentials flow requiring tenantId, clientId, and clientSecret global arguments. README now documents the correct auth mechanism, required role (Reader on subscriptions), and all required global arguments.

2026.08.14.4

Fixed: three issues surfaced after the #360 dedup fix. First, the ARM $filter=kind eq 'OpenAI' or kind eq 'AIServices' query param sent to the Cognitive Services list endpoint is not reliably enforced server-side, so non-token-emitting kinds (Face, ComputerVision, TextAnalytics, etc.) leaked into discovery and wasted metrics attempts; listAiResources now also filters client-side. Second, resources with zero usage in the lookback window were silently dropped with no log line at all, indistinguishable from a resource that was never attempted; scan_subscriptions now logs a debug line before each metrics attempt and an info line for zero-usage outcomes. Third, per-resource metrics failures logged only String(err), discarding detail needed to root-cause failures that don't reproduce via az rest; failure logs now include the resource's subscription/resourceGroup/kind and the raw error's name/message/stack.

2026.08.14.3

Fixed: listAiResources could enumerate the same ARM resource twice across paginated nextLink requests, causing scan_subscriptions to process more candidates than the deduplicated resource count reported by list_ai_resources — surfacing as some resources failing to scan and the result being marked truncated: true even though nothing was actually throttled or missing. Results are now deduped by ARM resource ID across pages. Per-page fetch counts and first/last resource IDs are now logged at debug level to make a repeat of this easier to diagnose from logs alone.

2026.08.14.1

Fixed: listAiResources ignored the ARM API's nextLink pagination cursor, silently dropping resources past the first page. GET .../Microsoft.CognitiveServices/accounts paginates at a small page size (observed: 2 per page); a subscription with 10 OpenAI/AIServices accounts was reporting only 1 in scan_subscriptions and list_ai_resources output. Both methods now follow nextLink until exhausted.

2026.07.31.1

Fixed: README incorrectly stated authentication uses az CLI (az login). The extension actually uses Azure AD client credentials flow requiring tenantId, clientId, and clientSecret global arguments. README now documents the correct auth mechanism, required role (Reader on subscriptions), and all required global arguments.

2026.08.14.3

Fixed: listAiResources could enumerate the same ARM resource twice across paginated nextLink requests, causing scan_subscriptions to process more candidates than the deduplicated resource count reported by list_ai_resources — surfacing as some resources failing to scan and the result being marked truncated: true even though nothing was actually throttled or missing. Results are now deduped by ARM resource ID across pages. Per-page fetch counts and first/last resource IDs are now logged at debug level to make a repeat of this easier to diagnose from logs alone.

2026.08.14.1

Fixed: listAiResources ignored the ARM API's nextLink pagination cursor, silently dropping resources past the first page. GET .../Microsoft.CognitiveServices/accounts paginates at a small page size (observed: 2 per page); a subscription with 10 OpenAI/AIServices accounts was reporting only 1 in scan_subscriptions and list_ai_resources output. Both methods now follow nextLink until exhausted.

2026.07.31.1

Fixed: README incorrectly stated authentication uses az CLI (az login). The extension actually uses Azure AD client credentials flow requiring tenantId, clientId, and clientSecret global arguments. README now documents the correct auth mechanism, required role (Reader on subscriptions), and all required global arguments.

2026.08.14.1

Fixed: listAiResources ignored the ARM API's nextLink pagination cursor, silently dropping resources past the first page. GET .../Microsoft.CognitiveServices/accounts paginates at a small page size (observed: 2 per page); a subscription with 10 OpenAI/AIServices accounts was reporting only 1 in scan_subscriptions and list_ai_resources output. Both methods now follow nextLink until exhausted.

2026.07.31.1

Fixed: README incorrectly stated authentication uses az CLI (az login). The extension actually uses Azure AD client credentials flow requiring tenantId, clientId, and clientSecret global arguments. README now documents the correct auth mechanism, required role (Reader on subscriptions), and all required global arguments.

2026.08.14.1

Fixed: listAiResources ignored the ARM API's nextLink pagination cursor, silently dropping resources past the first page. GET .../Microsoft.CognitiveServices/accounts paginates at a small page size (observed: 2 per page); a subscription with 10 OpenAI/AIServices accounts was reporting only 1 in scan_subscriptions and list_ai_resources output. Both methods now follow nextLink until exhausted.

2026.07.31.1

Fixed: README incorrectly stated authentication uses az CLI (az login). The extension actually uses Azure AD client credentials flow requiring tenantId, clientId, and clientSecret global arguments. README now documents the correct auth mechanism, required role (Reader on subscriptions), and all required global arguments.

2026.07.31.1

Fixed: README incorrectly stated authentication uses az CLI (az login). The extension actually uses Azure AD client credentials flow requiring tenantId, clientId, and clientSecret global arguments. README now documents the correct auth mechanism, required role (Reader on subscriptions), and all required global arguments.

2026.07.21.1

Changed: Authentication no longer shells out to az CLI. Auth now uses Azure AD client credentials flow (tenant_id + client_id + client_secret → access token). Resource discovery and metrics collection use the ARM REST API directly. This eliminates the az CLI runtime dependency.

Added: Three new required global arguments:

• tenantId (UUID) — Azure AD tenant ID
• clientId (UUID) — Azure AD application (client) ID
• clientSecret (sensitive) — Azure AD client secret

Upgrade note: This is a breaking change for existing model instances. You must add the three new global arguments:

swamp model create @webframp/azure/openai-usage azure-ai-usage \
  --global-arg 'subscriptions=["sub-id-1"]' \
  --global-arg 'tenantId=<your-tenant-id>' \
  --global-arg 'clientId=<your-app-id>' \
  --global-arg 'clientSecret=<vault:azure/sp-secret>'

The service principal needs Reader role on target subscriptions with Microsoft.CognitiveServices/accounts/read and Microsoft.Insights/metrics/read permissions.

2026.07.18.1

Added: An upgrades array entry (no-op) to openai_usage.ts for proper typeVersion tracking on existing instances. No schema or behavior changes.