> ## Documentation Index
> Fetch the complete documentation index at: https://docs.braintrust.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Structured outputs unsupported for Azure providers

> [!NOTE]
>
> **Applies to:**
>
> * Plan - Any
> * Deployment - Any
> *
> * Use case - Diagnose and resolve structured\_outputs errors when a model is served through an Azure/custom provider (example: claude-sonnet-5 via Azure Foundry)

## Summary

You see: "structured\_outputs not supported in your workspace" when using structured output. Likely cause: the model endpoint or custom provider (for example Azure Foundry) does not support the `structured_outputs` feature.

Fix: use text output or switch to a provider/model endpoint that supports structured outputs, or update provider configuration if supported.

## What is happening

The error is returned by the model provider or model endpoint, not the Braintrust workspace. Some custom provider deployments (notably Azure/custom provider setups) or certain model endpoints do not implement the structured\_outputs API feature. When the playground or API request requests structured output, the provider returns an invalid\_request\_error with that message.

Example minimal error payload:

```json theme={"theme":{"light":"github-light","dark":"github-dark-dimmed"}}
{
  "error": {
    "message": "structured_outputs not supported in your workspace",
    "type": "invalid_request_error",
    "request_id": "req_0000000000000000000000"
  }
}
```

This happens when:

* A model is routed through a custom provider (Azure Foundry, custom proxy, etc.). AND
* The provider/model does not advertise or implement structured output behavior.

## Fix or suggestion

### Option 1: Use text output (quick workaround)

* Switch the prompt/template output mode from Structured to Text in the playground or API request.
* Verify the model responds without the structured\_outputs error.
* Use downstream parsing on the text output if needed.

### Option 2: Change provider or model endpoint (preferred for structured data)

* Confirm whether the provider or model supports `structured_outputs` in its documentation or provider feature flags.
* If available, point the request to a provider or built-in Braintrust model that supports structured outputs.
* Reconfigure your custom provider to a model endpoint/version that advertises structured output support (if the provider offers that capability).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.