MCP OI-Wiki

Instructions

Implementation Reference

• main.py:8-17 (handler)

  The handler function for the MCP 'search' tool, decorated with @mcp.tool(). It takes a query string and delegates to the database search method.

  @mcp.tool()
  async def search(query: str):
      """
      OI Wiki 致力于成为一个免费开放且持续更新的编程竞赛知识整合站点，大家可以在这里获取与竞赛相关的、有趣又实用的知识。本工具能够在 OI-wiki 中搜索相关的知识点。
      
      query 应该比较详细，与要实现的算法/目标相关。比如：“求一个图的最小生成树”，“维护区间加/区间求和”
      
      @ param query 描述要实现的需求
      """
      return db.search(query)

• main.py:8-8 (registration)

  The @mcp.tool() decorator registers the 'search' function as an MCP tool.

  @mcp.tool()

• main.py:9-9 (schema)

  Function signature defining the input schema: query as str.

  async def search(query: str):

• database.py:33-46 (helper)

  The helper search method in OIWikiDB class that performs vector search using Milvus and retrieves the content of the most relevant OI-Wiki document.

  def search(self, query : str) :
      qvectors = list(self._embedding_model.embed([query]))
      results = self._client.search(
          collection_name=self._collection_name, 
          data=qvectors, 
          limit=1,
          output_fields=["path"]
      )
  
      path = os.path.join(self._docs_dir, results[0][0].entity.path)
      with open(path, 'r') as f:
          res = f.read()
  
      return res

• main.py:6-6 (helper)

  Instantiation of the database instance used by the search tool.

  db = OIWikiDB()