Sticking to the Ancient Greek theme we just released Athena, an AI assistant that helps users find products in a personalized way. Athena gathers context of what you're looking for, asks specific questions about the product category and budget, and then presents a few recommendations.
To power Athena we're using text embeddings of the name, description, and tag fields with SBERT. Additionally, we're using LangChain, GPT-3.5, GPUs on Digital Ocean's Paperspace, a Mongo database, and servers on AWS running self-hosted Typesense. Agora currently has a dataset of about 200 million products from stores on Shopify, WooCommerce, Squarespace, and Wix.
The vision for Athena is to use AI to help people find any product in the world. I'm continuing to build it for someone like myself: I want to shop from smaller stores around the world but unsure where to start. Still working out a lot of bugs and enhancements before letting Hermes (i.e our aptly named store crawler) index billions of e-commerce products. We'll also feed in more data like reviews, discussion threads, and automated user feedback to improve the recommendation experience.
I'd love any feedback!