
GoodsPoint combines a Rust/Axum backend with a React front end and MongoDB. Users can search by text image or voice and our AI refines queries to find precise industrial parts. The platform supports Hindi localization and a voice‑first interface so purchase orders can be spoken. A built‑in knowledge graph and LLM‑powered RAG Vector Search (compound-beta which internally uses a planner–executor model between llama 4 scout and llama 3.3 70b for routing and tool use) product recommendation engine is available to buyers. Voice is supported through the Web Speech API chained with an API Key free, almost ratelimit free relatively unknown Google Translate endpoint along with stored localisations allowing for the user to operate the website through Hindi or English. A knowledge graph is built based on product categories that supports lazy temporal decay and simple category relations (say, UnisexClothing is linked to MaleClothing and FemaleClothing) are used to optimise homepage product recommendations.
8 Jul 2025