Event-Driven Image Annotation and Retrieval System

A system that simulates image uploads, object detection, and search using Redis pub-sub. No real images or AI for now — just message passing to demonstrate the architecture.

How It Works

1. Script starts → automatically creates 2 fake images (cat.jpg, dog.png)
2. Processor "detects" objects in each image (simulated)
3. Storage saves the results
4. You search by keyword → system returns matching fake images

Message Flow (Redis Topics)

Upload → image.uploaded → Processor → embedding.done → Storage
                              ↓
Search → search.request → Search Handler → search.result → You

Run It

# Terminal 1
redis-server

# Terminal 2
python3 system.py

Try It

When you see Enter search term, type either:

• cat → returns the cat image
• dog → returns the dog image
• sofa → returns the cat image (because cat.jpg had 'sofa' as an object)
• ball → returns the dog image

Example Output

[1] UPLOAD: cat.jpg
[2] PROCESSOR: Analyzing cat.jpg
[3] PROCESSOR: Found objects -> ['cat', 'sofa']
[4] STORAGE: Saved (Total: 1 images)

Search> cat
RESULT: Found 1 image(s) matching 'cat'

What This Demonstrates

• Redis pub-sub messaging
• Asynchronous event-driven architecture
• Simulated AI inference and embedding
• Search by object labels
• Unit tests included (pytest test_system.py -v)

Files

• system.py – everything in one file
• test_system.py – unit tests