Effect 中文知识层的 MCP Server —— 让 AI 编码助手直接检索 Effect 官方文档的中文译文(234 页,带可核验引用;离线自包含、零依赖)
Six tools with complete input schemas and descriptions. Naming is clear and verb-driven (search_, get_, ask, glossary, translation_status, cite). Descriptions are substantive (80-180 chars), explaining WHAT and WHEN. Parameters are well-typed with enums (version in search_docs). However, output schemas are not documented in the source, LLMs cannot see what fields to expect from responses. Error handling is implicit (e.g., 'will clearly refuse' in ask tool) but not formalized with recovery guidance. All tools are READ_ONLY, reducing risk. Tool composition is sound: search_docs → get_page → ask forms a logical chain. Missing: explicit output schema documentation, error classification, and per-parameter constraints (e.g., limit bounds in search_docs).
就中文 Effect 文档提问。返回带引用的回答;若站内没有依据或该页尚未翻译,会明确拒答(citations 为空)。
解析并核验一条引用:给出规范化引用 ID、可解引用地址(/cite/<digest>.json)、原文片段、内容指纹、上游基线与官方原文地址。用途:独立核对某条引用是否仍成立、是否已漂移 —— 不要凭记忆判断引用是否过期。
取回某一页中文译文的完整 Markdown(含上游文件路径与基线 commit)。slug 形如 v4/getting-started/installation。
返回社区术语表:哪些术语保留英文、以及 CI 门禁禁止的译法。
在 effect-ts.cn 的**中文** Effect 文档中做关键词检索,返回最相关的页面与小节(含官方原文链接)。未翻译的官方页面也会被列出。
查询某页或全部页面的翻译状态(已翻译/未翻译、上游基线、是否落后)。不传 slug 则返回总体统计。
Output schemas not documented. LLMs cannot see what fields search_docs, get_page, ask, or cite return, forcing them to guess at response structure and risk extraction errors.
Parameter constraints missing. search_docs 'limit' has no min/max bounds; LLMs could pass 0, negative, or absurdly large values. glossary has empty properties object, unclear what it accepts.
Error handling not formalized. ask tool states it 'will clearly refuse' but no error schema or recovery guidance is documented. cite tool accepts multiple key formats but no validation or error cases are specified.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
get_page slug parameter lacks format guidance. Description says 'slug form like v4/getting-started/installation' but no regex pattern, length limit, or validation rules are stated. LLMs may pass invalid slugs.
glossary tool has no description of what it returns or when to call it. Users/LLMs cannot determine if it returns a list, a map, or a single object, or whether it should be called before translation work.