A lightweight pipeline for building and querying product-similarity search using text embeddings and approximate nearest-neighbor (ANN) indexing.
A toy implementation based on FAISS
• Exact vs HNSW ANN search • Latency–accuracy tradeoffs (efSearch) • 1.2M product embeddings (benchmarked) • Deployed interactive Streamlit demo on Hugging Face: https://huggingface.co/spaces/emdil99/product-search-faiss