trieve

Why we included this project

Teams building semantic search or retrieval-augmented generation over their own documents will find Trieve useful. The platform chunks content, indexes it with dense and sparse neural models, and exposes search, recommendations, filtering, grouping, and recency biasing through one API. You can supply your own embedding, re-ranking, or LLM providers rather than relying on the managed ones, and hybrid search with cross-encoder re-ranking is included. Self-hosting is a genuine option with guides for AWS, GCP, and Kubernetes, so teams that must keep data inside their own infrastructure are covered. The RAG routes assemble context and track topic-based memory, which cuts the work of wiring answer generation into an existing application.