ReAct agent with CLI, Streamlit web UI, and WeCom WebSocket bot entrypoints. Provides tools for RAG knowledge base retrieval, date queries, web search, and MCP tool integration.
Server has 5 tools with basic schemas and descriptions in Chinese. Naming follows verb_noun pattern (get_, bing_, rag_). However, descriptions are minimal (10-50 chars), lack WHEN_TO_USE guidance, and parameter descriptions are sparse. Schemas present but lack depth, no enums, ranges, or format constraints. Error handling and output schemas not visible in source. Tool composition is reasonable (search → crawl → summarize chain), but descriptions do not guide LLM selection or explain prerequisites.
抓取单个网页正文内容,返回可读纯文本与元数据。
使用必应中文搜索引擎检索网页,返回结构化 JSON 字符串。
获取当前日期字符串(按运行环境本地时区的日历日)。
扫描知识目录并增量更新向量索引(管理员或自动化任务使用)。
优先入口:检索本地向量知识库参考资料;工具只返回资料,不生成最终答案。
Descriptions are too brief (10 - 50 chars) and lack WHEN_TO_USE / WHEN_NOT_TO_USE guidance. LLMs cannot determine tool selection criteria.
Parameter descriptions are minimal or missing. 'query' in bing_search and rag_summarize lack format/length guidance. 'count' and 'offset' in bing_search lack range constraints (stated as 'default 10, max 50' but not in schema).
Output schemas not documented in source. LLMs cannot plan downstream tool calls or extract required fields (e.g., does bing_search return URLs, titles, snippets? Does rag_summarize return references or just text?).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
rag_rebuild_index is a WRITE operation but lacks confirmation/dry-run pattern. Agents could accidentally trigger expensive index rebuilds without safeguards.
No error handling guidance visible. Tools do not document retryable vs. fatal errors, or recovery steps (e.g., 'If bing_search returns 0 results, try a shorter query').