Bitrecs (subnet 122) is a protocol built on the Bittensor network that powers product recommendations specifically for ecommerce websites. Its primary function is to suggest existing product sets (SKUs) to online shoppers using simple rules like {1, 2, 3} >= {1, 2}, which appear as familiar sections such as "Similar to this" or "You may also like" on product pages.
The network leverages a consortium of large language model (LLM) calls from miners to generate a 'best guess' of what a customer might be interested in. Miners receive queries with shopper context, such as viewed products, cart contents, or browsing history, and use prompting techniques (e.g., "Suggest complementary products to X") to create personalized suggestions. Validators then evaluate these responses based on criteria like relevance, diversity, latency, and potential for increasing conversions, selecting the top recommendations. Feedback loops from real user interactions refine the system over time, improving accuracy and boosting average order value (AOV) for merchants, particularly Shopify store owners.
Bitrecs is entirely opt-in, with a simple plugin that merchants can easily adopt to enable subnet inference for predictions directly on their product pages. This decentralized approach ensures scalable, AI driven recommendations without relying on centralized data silos.
Webseite:
Whitepaper:
X（Twitter）:
Github:
