Show HN: Local Search Agent – offline RAG, no embeddings, free tier (github.com)

The newly introduced Local Search Agent is a Python framework that enables users to create an AI agent capable of searching and reasoning over local files without the need for cloud services or embeddings. By utilizing a BM25 keyword search through Meilisearch and a LangGraph agent loop, this tool empowers users to query their documents with full transparency, allowing retrieval of relevant information along with citations. This offline Retrieval-Augmented-Generation (RAG) system resolves common issues found in traditional RAG implementations, such as stale indexes and black-box retrieval, making it a robust alternative for handling sensitive data privately. The significance of this tool lies in its capacity to operate locally, alleviating concerns about data security and dependency on external APIs. Users can seamlessly integrate their document collections into the framework, which automatically indexes and manages them using structured metadata. The Local Search Agent supports multiple file formats, offers a user-friendly desktop interface, and includes features like role-based access control for multi-tenant usage. With this release, the AI/ML community gains a powerful and deterministic solution for local information retrieval, enhancing productivity and enabling more efficient workflows without the complexity of embedding technologies.