gorse-io/gorse

What it solves

Gorse provides a universal, open-source recommender system engine that allows online services to quickly integrate personalized recommendations. It eliminates the need to build a recommendation pipeline from scratch by automatically training models based on imported user, item, and interaction data.

How it works

Gorse is written in Go and operates as a single-node training and distributed prediction system. It uses a master node for model training and configuration, server nodes to expose RESTful APIs for real-time recommendations, and worker nodes to handle offline recommendations for users. It supports various databases for storage (MySQL, MongoDB, Postgres, ClickHouse) and uses Redis for caching intermediate results.

Who it’s for

Developers and service providers who want to add AI-powered recommendations to their applications without managing the complex underlying ML pipeline.

Highlights

• Multi-source recommendations: Supports collaborative filtering, item-to-item, user-to-user, and latest items.
• Multimodal support: Uses embeddings to handle text, images, and videos.
• AI-powered: Integrates both classical recommendation algorithms and LLM-based recommenders.
• Management GUI: Includes a dashboard for monitoring the system, managing data, and editing the recommendation pipeline.
• API-driven: Provides RESTful APIs for data CRUD operations and and recommendation requests.