Chat Deployment Issues Summary

Date: January 1, 2026
Status: Chat broken due to deployment failure

Root Cause

The latest deployment failed because:

  1. Deployment conflict: Previous deployment was still active (deployments triggered 33 seconds apart)
  2. Model Router still configured: Deployment parameters show azureAiModelRouter=model-router is still set

Deployment Failure Details

Error:

DeploymentActive: cannot be saved, because this would overwrite an existing deployment 
which is still active. The previous deployment was started at '1/1/2026 12:01:56 AM'

Deployment History:

  • 00:01:06Z - Previous deployment (SUCCESS)
  • 00:01:39Z - Latest deployment (FAILED - active deployment conflict)

Time between deployments: 33 seconds (violates 14-minute rule)

Configuration Issue

Model Router is still enabled:

azureAiModelRouter=model-router

This should be empty or not set to use direct model deployment.

Current Status

  • ✅ Code fix committed: GPT-5.1-chat API parameters fix is in codebase
  • ❌ Deployment failed: Active deployment conflict
  • ❌ Chat still broken: Fix not deployed
  • ⚠️ Model Router enabled: Should be disabled
  • ✅ Episodes/Sessions/Voice: Working (using older code)

Solution Steps

Step 1: Wait or Cancel Active Deployment

The active deployment will expire on 1/8/2026 12:01:56 AM if it doesn’t complete.

Option A: Wait

  • Wait for the active deployment to complete or fail
  • Then re-trigger deployment

Option B: Cancel (if possible)

  • Check Azure Portal for active deployments
  • Cancel if stuck

Step 2: Disable Model Router in Deployment

The deployment parameters need to be updated to remove Model Router:

Current (WRONG):

azureAiModelRouter=model-router

Should be:

azureAiModelRouter=  (empty)

This needs to be fixed in:

  • GitHub Actions workflow parameters
  • Bicep template parameters (if applicable)

Step 3: Wait for Deployment Window

IMPORTANT: Follow the 14-minute rule:

  • Wait at least 14 minutes between deployments
  • Check deployment status before triggering new one

Step 4: Re-trigger Deployment

Once the active deployment completes/fails and Model Router is disabled:

  1. Ensure code is committed
  2. Wait for deployment window (14 minutes)
  3. Trigger deployment (or push to trigger)
  4. Monitor deployment status
  5. Wait for completion

Step 5: Verify Configuration

After deployment succeeds, verify Azure Container Apps configuration:

az containerapp show \
  --name staging-env-api \
  --resource-group zimax-ai \
  --query "properties.template.containers[0].env[?name=='AZURE_AI_MODEL_ROUTER' || name=='AZURE_AI_API_VERSION' || name=='AZURE_AI_DEPLOYMENT']" \
  --output table

Expected:

  • AZURE_AI_MODEL_ROUTER = (not present or empty)
  • AZURE_AI_API_VERSION = 2024-12-01-preview
  • AZURE_AI_DEPLOYMENT = gpt-5.1-chat

Step 6: Test Chat

After successful deployment:

python3 scripts/test-chat-debug.py \
  --token "YOUR_JWT_TOKEN" \
  --message "hi"

Expected: HTTP 200 with actual agent response

Files to Update

1. GitHub Actions Workflow

Check .github/workflows/deploy.yml for:

azureAiModelRouter: model-router  # ❌ Should be empty

Change to:

azureAiModelRouter: ""  # ✅ Empty = use direct model

Or remove the parameter entirely if it’s optional.

2. Bicep Template (if applicable)

Check infra/main.bicep for Model Router parameter default.

Summary

  1. Deployment conflict - Wait for active deployment to complete
  2. Model Router enabled - Disable in deployment parameters
  3. Code fix ready - Already committed, just needs successful deployment
  4. Follow 14-minute rule - Don’t trigger deployments too quickly
  5. Verify after deploy - Check configuration and test chat endpoint

Once these issues are resolved and deployment succeeds, chat should work.

  • docs/troubleshooting/deployment-failure-chat-fix.md - Deployment failure details
  • docs/troubleshooting/gpt-5.1-chat-api-parameters-fix.md - Code fix details
  • .github/workflows/deploy.yml - Deployment workflow