A fashion product search engine powered by AI that enables both text-to-image and image-to-image search capabilities using CLIP for multimodal embeddings, Qdrant for vector storage and queries, and AWS S3 for image hosting; developed FastAPI backend, Dockerized it, and deployed on AWS ECS. Integrated with AWS S3-hosted frontend via AWS CloudFront, secured API with AWS API Gateway for HTTPS, enabling efficient product recommendations.

• Text-to-Image Search: Find fashion products using natural language descriptions
• Image-to-Image Search: Upload an image to find similar fashion items
• Vector Similarity Search: Powered by CLIP embeddings and Qdrant vector database
• Scalable Architecture: Load-balanced deployment with multiple app instances
• REST API: FastAPI-based API with automatic documentation

• Backend: FastAPI, Python 3.13
• AI/ML: OpenAI CLIP, PyTorch, Sentence Transformers
• Vector Database: Qdrant
• Containerization: Docker, Docker Compose
• Data Processing: Pandas, NumPy, Pillow

• Docker
• Python 3.13+ (for local development)

1. Clone the repository:

git clone https://github.com/Aayush621/Style-AI.git
cd style-ai

2. Create environment file:

cp .env.example .env

3. Configure your .env file:

QDRANT_URL=your_qdrant_url
QDRANT_API_KEY=your_qdrant_api_key
S3_BUCKET_BASE_URL=your_s3_bucket_url

# Access the API
# App instances: http://localhost:8000

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Run the API server
cd app
uvicorn api:app --reload --port 8000

Once running, access the interactive API documentation:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

GET /health

POST /search/text-to-image
Content-Type: application/json

{
  "query": "red summer dress",
  "top_n": 4
}

POST /search/image-to-image
Content-Type: multipart/form-data

{
  "image": [uploaded_file],
  "top_n": 4
}

search_ai/
├── app/                    # FastAPI application
│   ├── api.py             # Main API endpoints
│   ├── Dockerfile         # Container configuration
│   └── requirements.txt   # Python dependencies
├── data/                  # Dataset files
│   ├── image_paths_dl.txt # Image file paths
│   └── styles.csv         # Product metadata
├── embeddings/            # Pre-computed embeddings
│   ├── image_embeddings_aligned.npy
│   ├── text_embeddings_aligned.npy
│   └── product_ids_aligned.npy
└── qdrant_ingestion.py   # Data ingestion script

# Run the ingestion script to populate Qdrant
python qdrant_ingestion.py

This script:

• Loads fashion product data from CSV
• Processes images with CLIP model
• Stores embeddings in Qdrant vector database
• Aligns text and image embeddings

The system uses a microservices architecture:

• App Instance: FastAPI server (port 8000)
• Qdrant: Vector database for similarity search
• CLIP Model: Text and image embedding generation

1. Text Search:

   • User submits text query
   • CLIP generates text embedding
   • Qdrant finds similar image embeddings
   • Returns ranked fashion products

2. Image Search:

   • User uploads image
   • CLIP generates image embedding
   • Qdrant finds similar image embeddings
   • Returns visually similar fashion products

1. Fork the repository
2. Create feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open a Pull Request

This project is licensed under the MIT License - see the LICENSE file for details.

Never commit .env files containing sensitive credentials. Use .env.example for templates.