Testing AI Image Generation - Step by Step Guide

✅ Prerequisites

1. Environment variables set (already done):

   AZURE_IMAGE_API_KEY=DzQKmjEA8WofYwOxbLaPVmek5PWqMt3XeBUuDNm5BGVybFJwscQSJQQJ99CCACF24PCXJ3w3AAAAACOG1Joz
   AZURE_IMAGE_ENDPOINT=https://getmo-mn31nkpp-uaenorth.cognitiveservices.azure.com/openai/deployments/gpt-image-1.5/images/generations?api-version=2024-02-01

2. Services running:

   • PostgreSQL (port 5434)
   • Redis (port 6379)
   • DigitalOcean Spaces configured

🧪 Test Method 1: Quick Provider Test (2 minutes)

Test just the Azure connection and image generation:

# Run the test script
pnpm tsx scripts/test-image-generation.ts

Expected output:

🧪 Testing Azure GPT Image 1.5 Image Generation

📡 Step 1: Testing Azure provider connection...
✅ Provider initialized successfully

🎨 Step 2: Generating test image...
✅ Image generated successfully!
   Created at: 2026-04-24T...
   Image count: 1
   Has base64 data: true
   Base64 length: ~150000 chars

If this works, Azure GPT Image 1.5 is connected correctly! ✅

🔄 Test Method 2: End-to-End Flow (10 minutes)

Test the complete workflow from post creation to image generation:

Step 1: Start the Worker

# In terminal 1 - Start the agents worker
cd packages/agents
pnpm dev

Watch for:

[social-post-designer-worker] Worker started
[social-post-designer-worker] Listening on queue: agent__social-post-designer

Step 2: Start the API Server

# In terminal 2 - Start the API
cd apps/api
pnpm dev

Watch for:

Server listening at http://localhost:3003

Step 3: Start the Dashboard

# In terminal 3 - Start the DM dashboard
cd apps/dashboard
pnpm dev

Watch for:

Local: http://localhost:3000

Step 4: Create & Approve a Post

Option A: Via Dashboard (Recommended)

1. Open DM dashboard: http://localhost:3001 
2. Navigate to “Social Posts” or “Content”
3. Find a post in “DM Review” status
4. Click “Approve” button
5. Watch terminal 1 (worker logs) for image generation

Option B: Via API (curl)

# Get a post in dm_review status
curl http://localhost:3003/dm/v1/social \
  -H "Authorization: Bearer YOUR_DM_TOKEN"
 
# Approve the post (replace POST_ID)
curl -X POST http://localhost:3003/dm/v1/social/POST_ID/dm-approve \
  -H "Authorization: Bearer YOUR_DM_TOKEN" \
  -H "Content-Type: application/json"

Step 5: Watch the Worker Logs

In terminal 1, you should see:

[social-post-designer-worker] Job started: {...}
[social-post-designer-worker] Generating 1 images for social post...
[social-post-designer-worker] Reserved 2 credits for image generation
[social-post-designer-worker] Generating image 1/1...
[social-post-designer-worker] Image generated successfully
[social-post-designer-worker] Overlaying logo...
[social-post-designer-worker] Uploading to Spaces...
[social-post-designer-worker] Media record created: media-xxx
[social-post-designer-worker] Consumed 2 credits
[social-post-designer-worker] Post status updated to: client_review
[social-post-designer-worker] Job completed successfully

Step 6: Verify Images Created

Check in database:

# Run this in a terminal
pnpm tsx -e "
  import { db } from '@leadmetrics/db';
  const media = await db.media.findMany({
    where: { socialPostId: 'YOUR_POST_ID' },
    orderBy: { createdAt: 'desc' },
    take: 5
  });
  console.log(media);
"

Check in DigitalOcean Spaces:

• Navigate to your DO Spaces bucket
• Look for files under social-posters/ folder
• Images should be PNG files with tenant/post IDs in the path

Check in Dashboard:

• Refresh the post detail page
• Post status should now be “Client Review”
• Images should be visible in the preview

🎯 Test Method 3: Direct Worker Trigger (Advanced)

Manually trigger the worker for a specific post:

// test-trigger-worker.ts
import { enqueueSocialPostDesigner } from "@leadmetrics/queue";
 
await enqueueSocialPostDesigner({
  activityId: "your-activity-id",
  tenantId: "your-tenant-id",
  wakeReason: "new_task",
});
 
console.log("✅ Worker job enqueued");

Run:

pnpm tsx test-trigger-worker.ts

🐛 Troubleshooting

Error: “Neither Azure nor OpenAI credentials configured”

Fix: Check environment variables are loaded:

node -e "console.log(process.env.AZURE_IMAGE_API_KEY?.substring(0, 10))"

Should output: DzQKmjEA8W

If not, restart your terminal or source the .env file.

Error: “Azure OpenAI image generation failed: 401”

Fix: API key is invalid or expired. Verify in Azure portal.

Error: “Azure OpenAI image generation failed: 429”

Fix: Rate limit reached. Wait a few minutes and retry.

Worker doesn’t process job

Fix:

1. Check Redis is running: redis-cli ping (should return PONG)
2. Check worker is running: Look for “Worker started” log
3. Check queue status: Use BullBoard or Redis CLI

Images not appearing in Spaces

Fix:

1. Check DO_SPACES_* environment variables
2. Verify bucket permissions
3. Check worker logs for upload errors

📊 Success Criteria

✅ Test script runs without errors
✅ Provider generates test image successfully
✅ Worker processes approval and generates images
✅ Images uploaded to DigitalOcean Spaces
✅ Media records created in database
✅ Post status advances to “client_review”
✅ Credits consumed and logged

📝 Next Steps After Testing

1. Production deployment: Update production environment variables
2. Monitor credits: Check credit consumption in billing dashboard
3. Quality review: Review generated images for brand consistency
4. Performance: Monitor image generation time and worker queue depth
5. Cost tracking: Track Azure API usage and costs

🔗 Related Documentation

• AI Image Generation Feature
• Azure Migration Guide
• Provider Package README
• Social Post Designer Worker