Skip to content
Hacker News front page

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

MinishLab open-sourced Semble, a code-search tool for agents that combines Model2Vec embeddings, BM25, RRF fusion, and reranking; on a 63-repo benchmark, it used 98% fewer tokens than grep+read, reached 0.854 NDCG@10, and ran CPU queries in about 1.5 ms.

Why it matters: MinishLab's open-source Semble chains lightweight semantic embeddings, BM25 and reranking into one retrieval pipeline, scoring NDCG@10 0.854 across 63 code repos at about 1.5ms CPU per query.

Read the original ↗Export Markdown