Show HN: Retrievo – In-memory hybrid search for .NET AI agents (github.com)

Retrievo, a newly announced open-source search library for .NET, merges traditional BM25 lexical matching with modern vector similarity techniques using Reciprocal Rank Fusion (RRF). This in-memory solution allows developers to implement hybrid search capabilities without requiring external databases or servers, making it ideal for applications with up to 10,000 documents. Unique features include auto-embedding via Azure OpenAI, mutable indexes for dynamic data updates, and a command-line interface to streamline indexing and querying processes. The significance of Retrievo lies in its ability to enhance search results through a combination of keyword relevance and semantic understanding, addressing the limitations of both lexical and vector-only searches. By providing detailed score breakdowns and incorporating metadata filtering, Retrievo aims to improve the efficiency of search operations in local agent memory and edge computing scenarios. With plans to extend functionality for larger datasets through approximate nearest neighbor (ANN) support, Retrievo positions itself as a versatile tool for developers delving into hybrid AI and machine learning retrieval systems.