DHLAB MCP Server

Instructions

Implementation Reference

• src/dhlab_mcp/server.py:167-199 (handler)

  The handler function decorated with @mcp.tool(), implementing the find_collocations tool using dhlab.Corpus to compute collocations for a given word in a document identified by URN, returning JSON results.

  @mcp.tool()
  def find_collocations(
      urn: str,
      word: str,
      window: int = 5,
      limit: int = 100,
  ) -> str:
      """Find collocations (words that appear near the target word) in a document.
  
      Args:
          urn: URN identifier for the document
          word: Target word to find collocations for
          window: Size of context window (default: 5)
          limit: Maximum number of collocations to return (default: 100)
  
      Returns:
          JSON string containing collocation statistics
      """
      try:
          # Create corpus from URN
          corpus = dhlab.Corpus.from_identifiers([urn])
  
          if len(corpus.corpus) == 0:
              return f"No document found for URN: {urn}"
  
          # Get collocations using corpus method
          colls = corpus.coll(words=word, before=window, after=window)
  
          if colls.coll is not None and len(colls.coll) > 0:
              return colls.coll.to_json(orient='records', force_ascii=False)
          return "No collocations found"
      except Exception as e:
          return f"Error finding collocations: {str(e)}"

• src/dhlab_mcp/server.py:167-167 (registration)

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

  @mcp.tool()

• src/dhlab_mcp/server.py:168-173 (schema)

  Function signature defines the input schema (parameters with types and defaults) and output type for the tool.

  def find_collocations(
      urn: str,
      word: str,
      window: int = 5,
      limit: int = 100,
  ) -> str: