🛍️ Semantic Product Search Engine

A high-performance, multi-modal search engine built with CLIP and FAISS (HNSW). This project allows users to search for physical products using natural language descriptions (e.g., "blue running shoes") instead of just keywords.

🚀 Technical Highlights

• Model: OpenAI's CLIP (clip-ViT-B-32) used for multi-modal embeddings.
• Vector Database: FAISS (Facebook AI Similarity Search) using an HNSW (Hierarchical Navigable Small World) index for $O(\log N)$ search complexity.
• Backend: FastAPI for high-concurrency request handling.
• Frontend: Streamlit for a clean, interactive user experience.

🏗️ System Architecture

1. Ingestion: Images are processed through a Vision Transformer to create 512-dimension vectors.
2. Indexing: Vectors are inserted into a proximity graph (HNSW) to allow for sub-20ms retrieval.
3. Inference: Natural language queries are encoded into the same vector space, and a "Nearest Neighbor" search finds the most relevant images.

🛠️ Installation & Usage

1. pip install -r requirements.txt
2. python setup_data.py # Downloads professional dataset
3. python build_index.py # Generates vector embeddings
4. uvicorn api:app # Starts the backend
5. streamlit run app.py # Starts the UI