Run the agent

This is the main event. We:

1. Enter the DaytonaAsyncComputer context manager, which starts the sandbox's computer-use processes (Xvfb, window manager, VNC).
2. Start a session recording. sandbox.computer_use.recording.start(...) returns a handle we need later to stop and download.
3. Build the Agent with a single tool, ComputerTool(computer=computer), and run it with Runner.run(...) inside a trace(...) block so the run shows up in the OpenAI traces dashboard.
4. In a finally block, stop the recording and download it locally, using the filename Daytona reports (currently an .mp4).

max_turns=50 is a generous ceiling for a form this size; a good run will come in well under that.

async with DaytonaAsyncComputer(sandbox) as computer:
    recording = await sandbox.computer_use.recording.start("form-fill")
    print(f"Recording started: {recording.id}")

    try:
        with trace("Daytona form-fill demo"):
            agent = Agent(
                name="Form filler",
                instructions=INSTRUCTIONS,
                tools=[ComputerTool(computer=computer)],
                model="gpt-5.4",
            )
            result = await Runner.run(agent, TASK, max_turns=50)
            print(f"\n--- Final output ---\n{result.final_output}")
    finally:
        stopped = await sandbox.computer_use.recording.stop(recording.id)
        print(f"\nRecording stopped: {stopped.file_name} ({stopped.status})")

        local_recording_path = Path(stopped.file_name).name
        await sandbox.computer_use.recording.download(recording.id, local_recording_path)
        print(f"Recording downloaded to: {local_recording_path}")

Recording started: de5b8077-352e-41b0-b2da-165ede8cbca0

--- Final output ---
DONE

Recording stopped: de5b8077-352e-41b0-b2da-165ede8cbca0_form-fill_20260419_161419.mp4 (completed)
Recording downloaded to: de5b8077-352e-41b0-b2da-165ede8cbca0_form-fill_20260419_161419.mp4