{"slug": "show-hn-sweet-search-local-code-search-for-agents-5-33-cheaper-runs", "title": "Show HN: Sweet-search – local code search for agents (5–33% cheaper runs)", "summary": "Developer Marko released sweet-search, a local code search tool for Claude Code, Codex and Opencode, claiming it cuts agentic run costs by 5–33% with Sol 6.1 and Opus 5.5 while matching native search on solved tasks (481 of 1,000 pooled across 5 model-harness combinations on 200 held-out SWE-rebench tasks). The tool ships 6 tools including hybrid search with late interaction reranking and a custom grep, plus GPU-accelerated indexing and incremental indexing; Marko reports the custom MaxSim kernel runs 1.4–7.3× faster than NumPy and PyTorch for CPU reranking, the custom grep is 2.46× faster end-to-end than ripgrep, and the hybrid-search tool is state of the art on M2CRB and zero-shot state of the art on CoSQA.", "body_md": "Hi everyone :)\n\nAfter almost a year of fighting windmills, I am proud to say I built sweet-search, a local code search tool for Claude Code, Codex and Opencode.\n\nIt delivers 5-33% cheaper agentic runs with Sol 6.1 and Opus 5.5 with Claude Code, Codex and Opencode, with pooled solved tasks equal vs native (both 481 of 1,000 tasks pooled on 5 model*harness combos). Measured on 200 held-out SWE-rebench tasks. Detailed bench results are in the README.\n\nIt consists of 6 tools: hybrid search with late interaction reranking, my version of grep, and others. It also has GPU accelerated indexing (for embedding and late interaction models) and low footprint incremental indexing.\n\nSome highlights: 1. The ANN kernel is custom and very competitive with FAISS and USearch. 2. The custom MaxSim kernel is 1.4–7.3× faster than NumPy and PyTorch (reranking on CPU). 3. My hybrid-search tool is SOTA on M2CRB and zero-shot SOTA on CoSQA. 4. My custom grep tool is 2.46x faster e2e than ripgrep and ranks hits in a specific way (inspired by Sainte-Laguë).\n\nHere is the promo video I built with fframes and Opus 5.5: [https://www.youtube.com/watch?v=lJzgXRQprs4](https://www.youtube.com/watch?v=lJzgXRQprs4)\n\nWould love to hear your opinions, Marko :)\n\nComments URL: [https://news.ycombinator.com/item?id=50326635](https://news.ycombinator.com/item?id=50326635)\n\nPoints: 2\n\n# Comments: 0", "url": "https://wpnews.pro/news/show-hn-sweet-search-local-code-search-for-agents-5-33-cheaper-runs", "canonical_source": "https://github.com/mrsladoje/sweet-search", "published_at": "2026-11-01 15:36:02+00:00", "updated_at": "2026-11-01 15:49:45.186122+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "ai-search"], "entities": ["sweet-search", "Marko", "Claude Code", "Codex", "Opencode", "Sol 6.1", "Opus 5.5", "ripgrep"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/show-hn-sweet-search-local-code-search-for-agents-5-33-cheaper-runs", "markdown": "https://wpnews.pro/news/show-hn-sweet-search-local-code-search-for-agents-5-33-cheaper-runs.md", "text": "https://wpnews.pro/news/show-hn-sweet-search-local-code-search-for-agents-5-33-cheaper-runs.txt", "jsonld": "https://wpnews.pro/news/show-hn-sweet-search-local-code-search-for-agents-5-33-cheaper-runs.jsonld"}}