Retirement notice LZ_6-Y1Z (Azure OpenAI Assistants) shows no impacted resources

hi Mohamed Farook Ali Bahardeen, thx for sharing urs issue here at Q&A portal,

Service Health impact lists are often subscription-level and don’t always have resource-level detection for API usage.

The notice may flag a subscription because it has Azure OpenAI resources that could use Assistants, not because Microsoft found a specific active assistant resource. Assistants usage is API-level, so there may be nothing in ARM resource inventory that clearly says ‘this account is using Assistants’.

Empty Impacted Resources doesn’t mean u can ignore the notice. It just means the advisory can’t map the usage to a specific Azure resource.

Best way to find it is search ur app code, CI/CD vars, Key Vault secrets, API gateway logs, App Insights, and diagnostic logs for endpoints like

/openai/assistants

/openai/threads

/openai/runs

/openai/vector_stores

Search for SDK calls too, like AssistantsClient, create_assistant, threads.create, or old Azure OpenAI Assistants API versions.

If diagnostic logging was enabled, query Log Analytics/API Management/App Insights for request paths containing assistants, threads, or runs. If logs weren’t enabled, code/config search is basically ur best option.

So yeah, the subscription is flagged as potentially impacted. The portal isn’t saying ‘no impact’; it’s saying it can’t identify the exact resource for u.

rgds,

Alex

&

If my answer was helpful pls mark it and additional thx if u follow me at Q&A portal

Hello @Mohamed Farook Ali Bahardeen ,

Welcome to Microsoft Q&A .Thank you for reaching out to us.

The Service Health notification identifies the subscription as potentially associated with the Azure OpenAI Assistants API retirement; however, the advisory does not currently identify the specific workload or resource responsible for that association.

Because the notification does not currently expose the affected workload, additional validation is recommended through Azure OpenAI resource inventory, Assistants object checks, application analysis and telemetry review.

The following validation process will help determine whether the notification reflects active usage, historical usage or no identifiable Assistants API dependency, and will help guide any migration activities required before the retirement deadline.

1. Inventory Azure OpenAI Resources Begin by identifying Azure OpenAI resources within the impacted subscriptions.

      resources
      | where type =~ "microsoft.cognitiveservices/accounts"
      | where kind =~ "OpenAI"
      | project name,
                resourceGroup,
                subscriptionId,
                location,
                id
   

   This establishes the list of Azure OpenAI resources that should be reviewed further.
2. Reviewing Existing Assistants Objects For each Azure OpenAI endpoint, review whether Assistants objects currently exist.

      GET https://<resource-name>.openai.azure.com/openai/assistants?api-version=<supported-assistants-api-version>
   

   When reviewing the results:
   • Returned Assistant objects indicate that Assistants-related artifacts currently exist and should be reviewed as part of migration planning.
   • Appropriate Azure OpenAI data-plane permissions are required.
   • The API version used must be supported by the Azure OpenAI resource and region.
   • An empty response indicates that no Assistants objects are currently returned, but it does not completely rule out historical, transient, or application-managed usage.
3. Reviewing Application Code and Configuration Assistants API dependencies may exist within applications and may not be visible through the Azure portal. Review:
   • Source code repositories
   • Deployment pipelines
   • Infrastructure templates
   • Application configuration files
   /openai/assistants /openai/threads /openai/runs /openai/messages /openai/files client.beta.assistants client.beta.threads assistant_id thread_id run_id Particular attention needs to be given to:
   • Azure App Service applications
   • Azure Functions
   • Azure Kubernetes Service (AKS) workloads
   • Virtual Machines
   • Third-party or internally hosted applications integrated with Azure OpenAI
4. Reviewing Monitoring and Telemetry If diagnostic logging and application telemetry have been enabled, review activity related to Assistants API operations. Potential sources include:
   • Azure Monitor
   • Log Analytics
   • Application Insights
   • Azure API Management

      AzureDiagnostics
      | where requestUri_s contains "/openai/assistants"
         or requestUri_s contains "/openai/threads"
         or requestUri_s contains "/openai/runs"
      | project TimeGenerated,
                ResourceId,
                requestUri_s,
                OperationName
   

   Important considerations:
   • The diagnostic table and available fields may vary depending on the logging configuration.
   • Historical activity may not be available if diagnostic settings were not enabled before the API calls occurred.
5. Assessing Migration Requirements Once any Assistants API dependencies have been identified, assess whether migration activities are required. Workloads using Assistants API operations such as: /openai/assistants /openai/threads /openai/runs should be reviewed against the migration guidance. For workloads already using Azure AI Foundry Agent Service, migration requirements may differ depending on the implementation.

The following references might be helpful , please check them out

• What is Microsoft Foundry Agent Service? - Microsoft Foundry | Microsoft Learn
• Migrate to the new Foundry Agent Service - Microsoft Foundry | Microsoft Learn
• Diagnostic Settings in Azure Monitor - Azure Monitor | Microsoft Learn
• Application Insights OpenTelemetry observability overview - Azure Monitor | Microsoft Learn
• Overview of Azure Resource Graph - Azure Resource Graph | Microsoft Learn
• Azure OpenAI in Microsoft Foundry Models REST API reference - Microsoft Foundry | Microsoft Learn
• What is Azure Service Health? - Azure Service Health | Microsoft Learn

Please let us know if the response was helpful

Thank you