
Typegap
io.github.CloudCorpRecords由 GitHub 账号 CloudCorpRecords 发布v0.1.0更新于 Nov 3, 2026
按文件对类型债务进行排名:any 和 ts-ignore 按导入影响范围加权计数。
AI 翻译由 AI 模型根据下方的原文描述翻译,可能有误。
原文描述
Ranks type debt per file: any and ts-ignore counts weighted by import blast radius.
概览
Typegap 的本地 MCP 服务;无需登录;STDIO。
暂无 AI 概览:以下信息来自注册中心条目和服务本身。
安装
在 SourceWeft 中
- 打开 控制台中的 Typegap,将其添加到工作区。
- 为需要使用其工具的对话启用该服务。
Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。
其他 MCP 客户端
参照 仓库 中的启动说明。
README
typegap
Every codebase has files where the types gave up: a cast to
any here, a @ts-ignore there. typegap tells you which of
those files hurt most. It counts type debt per file and
weights each file's score by how many other files import
it, so a sloppy file that half the codebase depends on
ranks above an equally sloppy leaf file nobody touches.
-
The
anycount — six textual pattern families, applied line by line in a fixed order, with each family's matched spans removed before the next family runs so no span is counted twice:Array<any>(a boundary — a character that is not[A-Za-z0-9_$], or the line edge — is required beforeArray)Record<string, any>andRecord<string,any>(a boundary is required beforeRecord)<any>(no boundary requirement)any[](a boundary is required beforeany, which is whycompany[]does not count): any(the character afteranymust be a boundary)as any(a boundary is required beforeasand afterany, which is whyhas anydoes not count)
Because earlier families remove their spans,
Array<any>counts once rather than again as<any>, and: any[]counts once underany[]. -
Suppression counts — plain substring counts over the whole file:
@ts-ignore,@ts-expect-error, and@ts-nocheck. -
Debt and blast radius —
debt = 2 * any_count + 3 * ts_ignore_count + 1 * ts_expect_error_count + 5 * ts_nocheck_count, andblast_radius = debt * (1 + imported_by). Suppressions that silence the compiler weigh more than a strayany, and@ts-nocheck— which opts a whole file out — weighs most of all.
Counting is textual on purpose: comments and string literals are not parsed, so a pattern inside one counts like code. That keeps the tool deterministic and honest about its scope — it measures where debt markers cluster, not what the compiler thinks of them.
The result lists only the files with debt, sorted by blast
radius descending, then debt descending, then path
ascending. Alongside the list it reports how many files
were scanned, how many carry debt, the any and debt
totals over every supplied file, and the worst file's
path (null when nothing supplied carries debt).
Validation is strict: a missing or empty file list, a file
missing its path or content, an empty, non-string, or
duplicate path, a non-string content, or an
imported_by that is not an integer >= 0 is an error
naming the offending path (for example
files[1].imported_by must be an integer >= 0), and no
partial ranking is produced. The computation is pure and
local: no network, no configuration, no files read beyond
the input you pass in.
CLI
args.json contains the full arguments object:
imported_by is optional per file and defaults to 0;
content is required but may be empty. The ranking is
printed as JSON:
Exit codes: 0 on success, 1 on a runtime error (including
an unreadable or invalid input file, or invalid input), 2
on a usage error.
MCP server
MCP client config:
Tools
-
rank_type_debt— Rank files by type debt.files(array, required, at least one) — source files as{path, content, imported_by?}objects;pathis unique across the set,contentis the full source text, andimported_bysays how many other files import the file
Returns
{files, scanned_files, files_with_debt, total_any, total_debt, worst_file}.
Docker
Development
License
MIT — Copyright (c) 2026 Rene Turcios
来源:README.md,提交 0bbb2a8
工具
0版本历史
1- v0.1.0最新Nov 3, 2026
评分
已满足适用检查项的 50%
该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。
- 近 30 天内可连通25 分
请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。
- 已声明认证方式10 分
请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。
- 已列出工具15 分
请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。
- 描述必需15 分
- 图标10 分
- 文档15 分
- 许可证10 分