Coach AI

Instructions

Implementation Reference

• src/coach_ai/server.py:166-216 (handler)

  The parse_task_list tool is implemented as an async function in server.py, decorated with @mcp.tool(). It uses the parse_natural_language_task_list helper function to parse the text.

  @mcp.tool()
  async def parse_task_list(
      text: str,
      default_priority: str = "medium",
      default_timeframe: Optional[str] = None,
  ) -> str:
      """Parse natural language task list into structured todos (does NOT create them).
  
      Use this to preview what will be created before committing.
      Then use add_todo(todos_batch=...) to create them.
  
      Or use brain_dump_tasks() for a one-step workflow.
  
      Supports multiple formats:
      - Bullet points: "- Task 1"
      - Numbered lists: "1. Task 2"
      - Checkboxes: "[ ] Task 3"
      - Section headers: "Sprint work:\\n- Task 4"
  
      Smart extraction:
      - Priority: "#high", "URGENT:", "!!"
      - Timeframe: "this week", "(this month)", "someday"
      - Time: "(30min)", "~2h"
      - Energy: "[high energy]", "[low effort]"
      - Theme: "#sprint", "@admin"
  
      Args:
          text: Natural language task list
          default_priority: Default priority for tasks
          default_timeframe: Default timeframe for tasks
  
      Returns:
          JSON string of parsed todos
      """
      result = parse_natural_language_task_list(text, default_priority, default_timeframe)
  
      response = f"{result['parse_summary']}\n\n"
  
      if result['parsed_todos']:
          response += "Parsed todos (ready for add_todo batch mode):\n"
          response += json.dumps(result['parsed_todos'], indent=2)
  
      if result['unparseable_lines']:
          response += "\n\nCould not parse:\n"
          for line in result['unparseable_lines']:
              response += f"  - {line}\n"
  
      response += f"\n\nTo create these todos, use: add_todo(todos_batch='{json.dumps(result['parsed_todos'])}')"
  
      return response