快速混合代码搜索代理。 纯Go，单静态二进制，直接插入MCP，兼容MinishLab/semble——工具模式相同，输出格式相同，安装步骤切换到Go二进制文件。

ken 是 semble 的 Go 移植版：BM25 词汇 + Model2Vec 语义嵌入 + RRF 融合 + 代码感知的重新排序器，检索算法是从 semble 的 search.py + ranking/*.py 逐字移植而来。

为什么是肯

• ~97% recall@10默认（混合）模式 — 在 semble 的 1,251 查询基准测试中为 0.967 NL / 0.995 符号**，而 grep 为 ~99.9% — 同时使代理花费的token数**~46×比grep + Read少（NL 查询中为4,120对189,773个中位数标记——在同一默认混合模式下测量）。对于“找到回答此问题的区块”，在接近奇偶校验召回时，token的优势是1到2个数量级。（重现：docs/BENCH.md。）
• 单静态二进制。 纯 Go，无 cgo，冷启动无 Python 解释器，索引无 GIL。免费交叉编译至 Linux / macOS / Windows（amd64/arm64）。
• semble 的直接插入。 相同的 search / [find_related MCP 工具模式和相同的标记字串线路格式——交换 command: 路径，现有代理保持不变。

ken

ken is code search built for coding agents: a pure-Go, single-binary engine that runs a 16M-parameter Model2Vec model in-process, with no cgo, Python, vector database, or API key. On ContextBench (53 SWE-bench Pro tasks, 8k-token budget), its default hybrid mode covered 52% of the relevant files versus 29% for BM25 alone, and on tasks where a gold line was retrieved reached the first gold line in about 40% fewer tokens. Warm queries take a millisecond or two, and with KEN_ANN=flat-i8 the watched dense vector store uses about 80% less memory.

Fast hybrid code search for agents. Pure Go, single static binary, drop-in MCP-compatible with MinishLab/semble — same tool schemas, same output format, install steps swapped to a Go binary.

ken is a Go port of semble: BM25 lexical + Model2Vec semantic embeddings + RRF fusion + a code-aware reranker, with the retrieval algorithm ported verbatim from semble’s search.py + ranking/*.py.

Why ken

• ~97% recall@10 in the default (hybrid) mode — 0.967 NL / 0.995 symbol on semble’s 1,251-query benchmark, vs grep’s ~99.9% — while costing an agent ~46× fewer tokens than grep + Read (4,120 vs 189,773 median tokens on NL queries — measured in the same default hybrid mode). For “find the chunk that answers this,” that’s a 1–2 order-of-magnitude token win at near-parity recall. (Reproduce: docs/BENCH.md.)
• Single static binary. Pure Go, no cgo, no Python interpreter on cold start, no GIL on indexing. Cross-compiles to Linux / macOS / Windows (amd64/arm64) for free.
• Drop-in for semble. Same search / find_related MCP tool schemas and the same markdown-string wire format — swap the command: path and existing agents work unchanged.
• Local, CPU-only. Embedding inference, BM25, and fusion all run on the CPU. No API keys, no GPU, no vector DB, air-gapped friendly.