Multimodal Product Recommender

• Tech Stack: FastAPI, pgvector, PostgreSQL, React, Vite, Jina / Cohere embeddings
• Github URL: Project Link

A recommendation engine that searches a product catalogue by both text and image. Queries are embedded into vectors and matched against stored product embeddings using pgvector similarity search inside PostgreSQL.

The embedding layer is pluggable: it can call hosted providers such as Jina or Cohere, or fall back to a lightweight offline provider so the app runs without external API keys. A FastAPI backend serves recommendations to a React + Vite frontend.