Trieve Introduction

Trieve works by integrating multiple advanced search and AI technologies into a unified API. It uses SPLADE for neural full-text search, dense vector embeddings for semantic search, and combines these in a hybrid approach with cross-encoder re-ranking. Users can upload their data to Trieve, which then handles chunking, indexing, and embedding. The API provides endpoints for search, recommendations, and RAG, allowing developers to easily implement these features in their applications. Trieve also offers tools for relevance tuning, including date recency biasing, term frequency boosting, and merchandising capabilities. For RAG applications, it manages message history and provides generated outputs with citations.