Method check_action

webarena/agent/agent.py:162–199

(
        self, trajectory: Trajectory, intent: str, meta_data: dict[str, Any], target_action: str
    )

        return action

    def check_action(
        self, trajectory: Trajectory, intent: str, meta_data: dict[str, Any], target_action: str
    ) -> Action:
        prompt = self.prompt_constructor.construct(
            trajectory, intent, meta_data
        )
        lm_config = self.lm_config
        n = 0
        
        # agent will retry if the action is not parsed correctly
        while True:
            response = target_action
            force_prefix = self.prompt_constructor.instruction[
                "meta_data"
            ].get("force_prefix", "")
            response = f"{force_prefix}{response}"
            n += 1
            try:
                parsed_response = self.prompt_constructor.extract_action(
                    response
                )
                if self.action_set_tag in ["id_accessibility_tree", "id_html_tree", "id_html_nasc_tree"]:
                    action = create_id_based_action(parsed_response)
                elif self.action_set_tag == "playwright":
                    action = create_playwright_action(parsed_response)
                else:
                    raise ValueError(
                        f"Unknown action type {self.action_set_tag}"
                    )
                action["raw_prediction"] = response
                break
            except ActionParsingError as e:
                if n >= lm_config.gen_config["max_retry"]:
                    action = create_none_action()
                    action["raw_prediction"] = response
                    break

        return prompt, action
    
    def reset(self, test_config_file: str) -> None:
        pass