Method _execute_step

phone_agent/agent_ios.py:154–267

Execute a single step of the agent loop.

(
        self, user_prompt: str | None = None, is_first: bool = False
    )

        self._step_count = 0

    def _execute_step(
        self, user_prompt: str | None = None, is_first: bool = False
    ) -> StepResult:
        """Execute a single step of the agent loop."""
        self._step_count += 1

        # Capture current screen state
        screenshot = get_screenshot(
            wda_url=self.agent_config.wda_url,
            session_id=self.agent_config.session_id,
            device_id=self.agent_config.device_id,
        )
        current_app = get_current_app(
            wda_url=self.agent_config.wda_url, session_id=self.agent_config.session_id
        )

        # Build messages
        if is_first:
            self._context.append(
                MessageBuilder.create_system_message(self.agent_config.system_prompt)
            )

            screen_info = MessageBuilder.build_screen_info(current_app)
            text_content = f"{user_prompt}\n\n{screen_info}"

            self._context.append(
                MessageBuilder.create_user_message(
                    text=text_content, image_base64=screenshot.base64_data
                )
            )
        else:
            screen_info = MessageBuilder.build_screen_info(current_app)
            text_content = f"** Screen Info **\n\n{screen_info}"

            self._context.append(
                MessageBuilder.create_user_message(
                    text=text_content, image_base64=screenshot.base64_data
                )
            )

        # Get model response
        try:
            response = self.model_client.request(self._context)
        except Exception as e:
            if self.agent_config.verbose:
                traceback.print_exc()
            return StepResult(
                success=False,
                finished=True,
                action=None,
                thinking="",
                message=f"Model error: {e}",
            )

        # Parse action from response
        try:
            action = parse_action(response.action)
        except ValueError: