Can't store completions in Azure Foundry

Hi Rodrigo Maldonado,

Thanks for sharing the question.

In Azure AI Foundry, the reason you cannot store completions is usually related to how logging and storage are configured, rather than an issue with the model itself.

First, it is important to understand that completions are not stored automatically by default. Azure AI Foundry only keeps request and response data if logging or tracing is explicitly enabled. If logging is not turned on, the completion is generated and returned to the application, but it is not saved anywhere for later viewing.

Please check whether tracing or logging is enabled for your Foundry project. Only runs that are tracked through the Foundry project pipeline can be stored and reviewed later. If the request is made directly to the model endpoint without project‑level tracking, the completion will not appear in storage or history. [learn.microsoft.com]

Another common cause is permissions. The project’s managed identity must have permission to write logs or artifacts to the connected storage account. If the identity has read access but not write access, the completion will run successfully but nothing will be saved.

Also verify that you are using a supported workflow. Storing completions works when:

• The request is executed inside an AI Foundry project
• Tracing is enabled
• A backing storage account is correctly connected to the project

If you are testing the model through a quick test or direct API call, those runs are usually ephemeral and are not persisted.

As a quick check, try running the same prompt through a tracked Foundry run and confirm that tracing is enabled. If it shows up there, then the issue is related to how the request is being sent rather than a problem with the service.

You can find more details in the official documentation here:

• Set up tracing in Azure AI Foundry (projects and runs) https://learn.microsoft.com/en-us/azure/foundry/observability/how-to/trace-agent-setup
• Enable diagnostic logging for Azure AI services
  https://learn.microsoft.com/en-us/azure/ai-services/diagnostic-logging

In short, completions can only be stored when they are executed as part of a tracked Foundry project with logging enabled and correct permissions. Once those pieces are in place, the outputs should be saved as expected.

Thank you!

Hi Rodrigo,

Try to replace the deployment name and resource name:

*curl "https://YOUR-RESOURCE-NAME.cognitiveservices.azure.com/openai/v1/chat/completions?api-version=2025-02-01-preview" *

  *-H "Content-Type: application/json" *

  *-H "api-key: $AZURE_OPENAI_API_KEY" *

  -d '{

    "model": "YOUR-DEPLOYMENT-NAME",

    "store": true,

    "metadata": {

      "user": "test-user",

      "category": "curl-test"

    },

    "messages": [

      { "role": "system", "content": "You are a helpful assistant." },

      { "role": "user", "content": "Hello!" }

    ]

  }'

I hope you find this helpful

♠ ♠ Ace ♠ ♠

Hi @Ace Jerraldven C. Enoc , I've tried this curl but it's not working for me

curl https://rodri-mnhynhh1-eastus2.cognitiveservices.azure.com/openai/v1/chat/completions?api-version=2025-02-01-preview \
  -H "Content-Type: application/json" \
  -H "api-key: $AZURE_OPENAI_API_KEY" \
  -d '{
    "model": "gpt-5.4-nano",
    "store": true,
	"metadata": {
      "user": "test-user",
      "category": "curl-test"
    },
    "messages": [
      {
        "role": "system",
        "content": "You are a helpful assistant."
      },
      {
        "role": "user",
        "content": "Hello!"
      }
    ]
  }'

This is the error I'm getting

{"error":{"code":"BadRequest","message":"API version not supported"}}