Analyze a Computer Use API Response

Objective

In this lab, you'll write a Python function that parses Computer Use API responses and extracts key information about the agent's actions.

Background

When Claude uses the Computer Use tool, it returns structured JSON describing actions like mouse clicks, keyboard input, and screenshots. Understanding this structure is essential for building monitoring and debugging systems.

Requirements

Create a function parse_computer_use_response(response: dict) -> dict that:

1. Extracts action type: Identify whether it's a mouse_move, left_click, type, key, screenshot, scroll, or wait action

2. Extracts coordinates: For mouse actions, return {"x": int, "y": int}

3. Extracts text content: For type actions, return the typed text

4. Identifies tool version: Extract computer_20250124 or older version

5. Returns structured output:

   {
       "action_type": str,
       "coordinates": {"x": int, "y": int} | None,
       "text": str | None,
       "tool_version": str,
       "is_screenshot_request": bool
   }
   

Example Input

response = {
    "type": "tool_use",
    "name": "computer",
    "input": {
        "action": "left_click",
        "coordinate": [512, 384]
    }
}

Example Output

{
    "action_type": "left_click",
    "coordinates": {"x": 512, "y": 384},
    "text": None,
    "tool_version": "computer_20250124",
    "is_screenshot_request": False
}

Hints

• The coordinate field is a list [x, y], not a dict
• Screenshot actions have action: "screenshot"
• The type action has a text field with the content to type
• Handle missing fields gracefully with defaults