Azure AI Foundry 400 Bad Request - Investigation

Status: 🔍 INVESTIGATING

Date: 2025-12-31
Issue: Azure AI Foundry API returns 400 Bad Request when calling model-router endpoint

Current Error 

FoundryChatClient: Error calling LLM: Client error '400 Bad Request' 
for url 'https://zimax-gw.azure-api.net/zimax/openai/deployments/model-router/chat/completions?api-version=2024-10-01-preview'

Endpoint Details

  • URL: https://zimax-gw.azure-api.net/zimax/openai/deployments/model-router/chat/completions?api-version=2024-10-01-preview
  • Type: APIM Gateway (not direct Azure AI Foundry)
  • Format: Azure AI Foundry format (/openai/deployments/{deployment}/chat/completions)
  • API Version: 2024-10-01-preview
  • Deployment: model-router

Code Configuration

From backend/agents/base.py:

# Endpoint detection
is_azure_openai = "openai.azure.com" in endpoint
is_apim_gateway = "/openai/v1" in endpoint  # This is FALSE for this endpoint

# URL construction
if not is_azure_openai and not is_apim_gateway:
    # Azure AI Foundry format
    self.url = f"{base}/openai/deployments/{deployment}/chat/completions?api-version={api_version}"
    self.is_openai_compat = False
    self.model = None

# Payload construction
payload = {
    "messages": [self._format_message(m) for m in messages],
}
# NO model parameter (is_openai_compat=False)
# Adds temperature and max_tokens (is_openai_compat=False)
if not self.is_openai_compat:
    payload["temperature"] = self.temperature
    payload["max_tokens"] = self.max_tokens

Possible Issues

1. APIM Gateway Expects OpenAI-Compatible Format

Hypothesis: APIM Gateway might expect OpenAI-compatible format even though URL suggests Foundry format.

Test: Try setting is_openai_compat=True for APIM Gateway endpoints.

2. Missing Required Fields

Hypothesis: APIM Gateway might require additional fields in the request.

Possible Missing Fields:

  • model parameter (even for Foundry format)
  • Different API version
  • Additional headers

3. Model Router Configuration

Hypothesis: The model-router deployment might not be configured correctly in APIM Gateway.

Check: Verify model-router deployment exists and is accessible.

4. Authentication Issues

Hypothesis: Bearer token or API key might be invalid or missing required scopes.

Check: Verify authentication headers are correct.

Enhanced Error Logging

We’ve added enhanced error logging to capture:

  • Error response body from API
  • Request URL
  • Payload structure
  • Message count

Next Step: After deployment completes, trigger a chat request and check logs for:

FoundryChatClient: Error response body: {actual error message}
FoundryChatClient: Request payload keys: [...]
FoundryChatClient: Payload messages count: X

Investigation Steps

  1. ✅ Enhanced Error Logging - Added to capture error response body
  2. ⏳ Wait for Deployment - Latest code with enhanced logging
  3. ⏳ Trigger Chat Request - Get actual error message from API
  4. ⏳ Analyze Error Response - Determine exact issue
  5. ⏳ Fix Request Format - Update payload/headers based on error
  6. ⏳ Test Again - Verify fix works
  • backend/agents/base.py - FoundryChatClient implementation
  • backend/core/config.py - Azure AI configuration
  • docs/troubleshooting/chat-endpoint-errors.md - Related errors

Status: Waiting for enhanced error logging to capture actual error message from Azure AI Foundry API