GitHub Repository

A framework that replace traditional RAG pipelines. Ingest any number of documents in multiple workspaces (channels, departments, etc.), index it with BM25, and let the agent search, fetch, and reason over it, exactly like searching the web, but entirely on your machine. No vector store, no embedding needed.

7 starsPython

Local Search Agent – RAG replacement, no embeddings, free tier

Name: Local Search Agent – RAG replacement, no embeddings, free tier
Availability: InStock
Author: Wissam-Metawee

by Wissam-Metawee·Jul 16, 2026·2 points·0 comments

Visit Project View on HN

AI Analysis

●●SolidBig BrainSolve My Problem

BM25 search beats vector drift when you need to know exactly why a doc was retrieved.

Strengths

•Meilisearch integration provides deterministic, debuggable retrieval unlike black-box embeddings.
•LangGraph agent loop autonomously searches, fetches, and cites sources without external APIs.
•Zero vector database overhead means no stale indexes or complex infrastructure to maintain.

Weaknesses

•Keyword-only search misses semantic connections that vector embeddings naturally capture.
•Requires running a local Meilisearch instance, adding a dependency for simple setups.