PlanWriterCtrlFlow.py

import json
from copy import deepcopy
from typing import Any, Dict, List

from flow_modules.aiflows.ChatFlowModule import ChatAtomicFlow

from dataclasses import dataclass


@dataclass
class Command:
    name: str
    description: str
    input_args: List[str]

class PlanWriterCtrlFlow(ChatAtomicFlow):
    """Refer to: https://huggingface.co/aiflows/JarvisFlowModule/blob/main/Controller_JarvisFlow.py
    This flow is a controller flow that controls the PlanWriterFlow.

    *Input Interface Non Initialized*:
    - `goal`: str

    *Input Interface Initialized*:
    - `feedback`: str
    - `goal`: str
    - `plan`: str

    *Output Interface*:
    - `command`: str
    - `command_args`: Dict[str, Any]

    *Configuration Parameters*:
    - `input_interface_non_initialized`: List[str] = ["goal"]
    - `input_interface_initialized`: List[str] = ["feedback", "goal", "plan"]
    - `output_interface`: List[str] = ["command", "command_args"]
    - `backend`: Dict[str, Any] : backend of the LLM
    - `commands`: List[Dict[str, Any]] : commands that the LLM can execute
    - `system_message_prompt_template`: str : the template of the system message prompt
    - `init_human_message_prompt_template`: str : the template of the initial human message prompt
    - `human_message_prompt_template`: str : the template of the human message prompt
    - `previous_messages`: Dict[str, Any] : the previous messages of the conversation (sliding window)

    """

    def __init__(
            self,
            commands: List[Command],
            **kwargs):
        """
        This function initializes the flow.
        :param commands: List[Command] : commands that the LLM can execute
        :type commands: List[Command]
        :param kwargs: other parameters
        :type kwargs: Dict[str, Any]
        """
        super().__init__(**kwargs)
        self.system_message_prompt_template = self.system_message_prompt_template.partial(
            commands=self._build_commands_manual(commands),
        )
        self.hint_for_model = """
        Make sure your response is in the following format:
              Response Format:
              {
              "command": "call plan writer, or to finish with a summary",
              "command_args": {
                  "arg name": "value"
                  }
              }
        """

    @staticmethod
    def _build_commands_manual(commands: List[Command]) -> str:
        """
        This function builds the manual for the commands.
        :param commands: List[Command] : commands that the LLM can execute
        :type commands: List[Command]
        :return: the manual for the commands
        :rtype: str
        """
        ret = ""
        for i, command in enumerate(commands):
            command_input_json_schema = json.dumps(
                {input_arg: f"YOUR_{input_arg.upper()}" for input_arg in command.input_args})
            ret += f"{i + 1}. {command.name}: {command.description} Input arguments (given in the JSON schema): {command_input_json_schema}\n"
        return ret

    @classmethod
    def instantiate_from_config(cls, config):
        """
        This function instantiates the flow from the config.
        :param config: the config of the flow
        :type config: Dict[str, Any]
        :return: the instantiated flow
        :rtype: ChatAtomicFlow
        """
        flow_config = deepcopy(config)

        kwargs = {"flow_config": flow_config}

        # ~~~ Set up prompts ~~~
        kwargs.update(cls._set_up_prompts(flow_config))

        # ~~~Set up backend ~~~
        kwargs.update(cls._set_up_backend(flow_config))

        # ~~~ Set up commands ~~~
        commands = flow_config["commands"]
        commands = [
            Command(name, command_conf["description"], command_conf["input_args"]) for name, command_conf in
            commands.items()
        ]
        kwargs.update({"commands": commands})

        # ~~~ Instantiate flow ~~~
        return cls(**kwargs)

    def _update_prompts_and_input(self, input_data: Dict[str, Any]):
        """
        This function updates the prompts and input data.
        :param input_data: the input data
        :type input_data: Dict[str, Any]
        :return: None
        :rtype: None
        """
        if 'goal' in input_data:
            input_data['goal'] += self.hint_for_model
        if 'feedback' in input_data:
            input_data['feedback'] += self.hint_for_model

    def run(self, input_data: Dict[str, Any]) -> Dict[str, Any]:
        """
        This function runs the flow.
        :param input_data: the input data
        :type input_data: Dict[str, Any]
        :return: the output of the flow
        :rtype: Dict[str, Any]
        """
        self._update_prompts_and_input(input_data)
        
        # ~~~when conversation is initialized, append the updated system prompts to the chat history ~~~
        if self._is_conversation_initialized():
            updated_system_message_content = self._get_message(self.system_message_prompt_template, input_data)
            self._state_update_add_chat_message(content=updated_system_message_content,
                                                role=self.flow_config["system_name"])
        
        while True:
            api_output = super().run(input_data)["api_output"].strip()
            try:
                response = json.loads(api_output)
                return response
            except (json.decoder.JSONDecodeError, json.JSONDecodeError):
                updated_system_message_content = self._get_message(self.system_message_prompt_template, input_data)
                self._state_update_add_chat_message(content=updated_system_message_content,
                                                    role=self.flow_config["system_name"])
                new_goal = "The previous respond cannot be parsed with json.loads. Next time, do not provide any comments or code blocks. Make sure your next response is purely json parsable."
                new_input_data = input_data.copy()
                new_input_data['feedback'] = new_goal
                input_data = new_input_data