Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep

Category: library

Tags: code-search, agent-tools, semantic-search

Score: 8.0/10 (Innovation: 8, Technical: 8, Documentation: 8, Utility: 8)

Semble is a code search library for AI agents that combines BM25 and lightweight static embeddings with code-aware reranking to deliver fast, token-efficient semantic code search entirely on CPU. Its innovative approach of using static models and hybrid retrieval achieves near-transformer accuracy with dramatically lower latency and resource usage, making it a practical tool for improving agent efficiency in code exploration.

Target audience: backend devs, ai engineers, devops

Repository: https://github.com/MinishLab/semble · Python · MIT · 535 stars

View on Hacker News