Skip to content

Releases: Asgoresmilelong/officegrep

v0.1.3 — zero-command first run

Choose a tag to compare

@Asgoresmilelong Asgoresmilelong released this 18 Oct 13:06

Zero-command first run.

When the index is empty, officegrep mcp now scans ~/Documents, ~/Downloads and ~/Desktop in the background (25 MB per-file cap). A search issued while that is still running answers with a "still building" note instead of a misleading "no results".

  • OFFICEGREP_AUTOINDEX=0 opts out
  • OFFICEGREP_ROOTS="a;b" scans other folders
  • 28 unit tests, CI on Windows + Linux, Python 3.9–3.12

Install: pip install https://github.com/Asgoresmilelong/officegrep/releases/download/v0.1.3/officegrep-0.1.3-py3-none-any.whl

v0.1.2 — watch mode

Choose a tag to compare

@Asgoresmilelong Asgoresmilelong released this 18 Oct 12:42

Adds officegrep watch: a dependency-free poller that keeps the index in sync as files change.

  • officegrep watch <paths> [--interval 2] [--once]
  • OFFICEGREP_WATCH="dir1;dir2" officegrep mcp starts the watcher inside the MCP server
  • SQLite now runs in WAL mode with a 5s busy timeout, so a watch process and an MCP server can share one index
  • BOM is stripped from text files (PowerShell/Notepad love adding it)
  • 23 unit tests, including a concurrent writer/reader test

Install: pip install https://github.com/Asgoresmilelong/officegrep/releases/download/v0.1.2/officegrep-0.1.2-py3-none-any.whl

v0.1.1 — MCPB bundle

Choose a tag to compare

@Asgoresmilelong Asgoresmilelong released this 18 Oct 07:59

Adds an MCPB bundle so the server can be installed by MCP hosts without cloning, plus server.json for the official MCP Registry.

v0.1.0

Choose a tag to compare

@Asgoresmilelong Asgoresmilelong released this 18 Oct 07:47

首个可用版本。

  • CLI:index / search / stats / mcp
  • MCP 服务器:手写 JSON-RPC over stdio,零 SDK 依赖,暴露 search_docs / list_docs / index_folder
  • 支持 docx / pptx / xlsx / pdf(可选 pypdf)/ txt / md
  • SQLite FTS5 trigram 索引:中英文子串匹配,无需分词器;bm25() 排序
  • 1–2 字查询回退扫描 + 按命中次数排序
  • 增量索引:(size, mtime) 快判 + SHA-1 兜底
  • 18 个单元测试,夹具现场生成
  • 实测:43 个文件 / 2260 块 / 2.16 s / 1.9 MiB 数据库