MCP server providing LLMs with real-time Spring documentation using hybrid RAG (vector + keyword search)
Server has 4 tools with mostly complete schemas and descriptions. Two tools are named identically (search_spring_docs), creating ambiguity. Descriptions are adequate (100-200 chars) but lack actionable guidance on when to use each variant. Input schemas are well-formed with proper types, enums, and constraints. However, output schemas are not documented in the provided code, and error handling guidance is absent. Tool composition is reasonable (single responsibility), but the duplicate naming violates clarity principles.
Get EverSpring MCP server status. Returns information about index health, available modules, and document counts. NOTE: First call takes ~12s to initialize.
List available Spring modules for filtering searches.
Searches the Spring ecosystem documentation using a Hybrid RAG (Vector + Keyword) approach. Returns exact code snippets and architectural explanations.
Search Spring documentation using hybrid retrieval. Uses cosine similarity + BM25 with Reciprocal Rank Fusion to find the most relevant documentation chunks. NOTE: First search takes ~12s to load the embedding model. Subsequent searches are fast (<1s).
Duplicate tool name 'search_spring_docs' with different schemas and descriptions. LLMs cannot distinguish between them; both will appear in tool selection, causing confusion and incorrect invocations.
Output schemas not documented in source code. LLMs cannot plan downstream operations or extract required fields (e.g., module, version, score) from search results.
No error handling guidance in tool descriptions. If a search returns no results or the embedding model fails to load, the LLM has no recovery path.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | <=2025-11-25 | v2 |
Parameter descriptions mention '~12s to load embedding model' but do not explain retry strategy or timeout behavior. LLMs may not know whether to retry or escalate.
No pagination support documented. If search results exceed context limits, tool provides no mechanism to fetch additional results or limit output size.