ChatDev/camel/agents/role_playing.py
2024-01-25 10:10:15 +08:00

280 lines
13 KiB
Python

# =========== Copyright 2023 @ CAMEL-AI.org. All Rights Reserved. ===========
# Licensed under the Apache License, Version 2.0 (the “License”);
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an “AS IS” BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# =========== Copyright 2023 @ CAMEL-AI.org. All Rights Reserved. ===========
import copy
from typing import Dict, List, Optional, Sequence, Tuple
from camel.agents import (
ChatAgent,
TaskPlannerAgent,
TaskSpecifyAgent,
)
from camel.agents.chat_agent import ChatAgentResponse
from camel.messages import ChatMessage, UserChatMessage
from camel.messages import SystemMessage
from camel.typing import ModelType, RoleType, TaskType, PhaseType
from chatdev.utils import log_arguments, log_visualize
@log_arguments
class RolePlaying:
r"""Role playing between two agents.
Args:
assistant_role_name (str): The name of the role played by the
assistant.
user_role_name (str): The name of the role played by the user.
critic_role_name (str): The name of the role played by the critic.
(default: :obj:`"critic"`)
task_prompt (str, optional): A prompt for the task to be performed.
(default: :obj:`""`)
with_task_specify (bool, optional): Whether to use a task specify
agent. (default: :obj:`True`)
with_task_planner (bool, optional): Whether to use a task planner
agent. (default: :obj:`False`)
with_critic_in_the_loop (bool, optional): Whether to include a critic
in the loop. (default: :obj:`False`)
model_type (ModelType, optional): The type of backend model to use.
(default: :obj:`ModelType.GPT_3_5_TURBO`)
task_type (TaskType, optional): The type of task to perform.
(default: :obj:`TaskType.AI_SOCIETY`)
assistant_agent_kwargs (Dict, optional): Additional arguments to pass
to the assistant agent. (default: :obj:`None`)
user_agent_kwargs (Dict, optional): Additional arguments to pass to
the user agent. (default: :obj:`None`)
task_specify_agent_kwargs (Dict, optional): Additional arguments to
pass to the task specify agent. (default: :obj:`None`)
task_planner_agent_kwargs (Dict, optional): Additional arguments to
pass to the task planner agent. (default: :obj:`None`)
critic_kwargs (Dict, optional): Additional arguments to pass to the
critic. (default: :obj:`None`)
sys_msg_generator_kwargs (Dict, optional): Additional arguments to
pass to the system message generator. (default: :obj:`None`)
extend_sys_msg_meta_dicts (List[Dict], optional): A list of dicts to
extend the system message meta dicts with. (default: :obj:`None`)
extend_task_specify_meta_dict (Dict, optional): A dict to extend the
task specify meta dict with. (default: :obj:`None`)
"""
def __init__(
self,
assistant_role_name: str,
user_role_name: str,
critic_role_name: str = "critic",
task_prompt: str = "",
assistant_role_prompt: str = "",
user_role_prompt: str = "",
user_role_type: Optional[RoleType] = None,
assistant_role_type: Optional[RoleType] = None,
with_task_specify: bool = True,
with_task_planner: bool = False,
with_critic_in_the_loop: bool = False,
critic_criteria: Optional[str] = None,
model_type: ModelType = ModelType.GPT_3_5_TURBO,
task_type: TaskType = TaskType.AI_SOCIETY,
assistant_agent_kwargs: Optional[Dict] = None,
user_agent_kwargs: Optional[Dict] = None,
task_specify_agent_kwargs: Optional[Dict] = None,
task_planner_agent_kwargs: Optional[Dict] = None,
critic_kwargs: Optional[Dict] = None,
sys_msg_generator_kwargs: Optional[Dict] = None,
extend_sys_msg_meta_dicts: Optional[List[Dict]] = None,
extend_task_specify_meta_dict: Optional[Dict] = None,
background_prompt: Optional[str] = "",
memory = None,
) -> None:
self.with_task_specify = with_task_specify
self.with_task_planner = with_task_planner
self.with_critic_in_the_loop = with_critic_in_the_loop
self.model_type = model_type
self.task_type = task_type
self.memory = memory
if with_task_specify:
task_specify_meta_dict = dict()
if self.task_type in [TaskType.AI_SOCIETY, TaskType.MISALIGNMENT]:
task_specify_meta_dict.update(
dict(assistant_role=assistant_role_name,
user_role=user_role_name))
if extend_task_specify_meta_dict is not None:
task_specify_meta_dict.update(extend_task_specify_meta_dict)
task_specify_agent = TaskSpecifyAgent(
self.model_type,
task_type=self.task_type,
**(task_specify_agent_kwargs or {}),
)
self.specified_task_prompt = task_specify_agent.step(
task_prompt,
meta_dict=task_specify_meta_dict,
)
task_prompt = self.specified_task_prompt
else:
self.specified_task_prompt = None
if with_task_planner:
task_planner_agent = TaskPlannerAgent(
self.model_type,
**(task_planner_agent_kwargs or {}),
)
self.planned_task_prompt = task_planner_agent.step(task_prompt)
task_prompt = f"{task_prompt}\n{self.planned_task_prompt}"
else:
self.planned_task_prompt = None
self.task_prompt = task_prompt
sys_msg_meta_dicts = [dict(chatdev_prompt=background_prompt, task=task_prompt)] * 2
if (extend_sys_msg_meta_dicts is None and self.task_type in [TaskType.AI_SOCIETY, TaskType.MISALIGNMENT,
TaskType.CHATDEV]):
extend_sys_msg_meta_dicts = [dict(assistant_role=assistant_role_name, user_role=user_role_name)] * 2
if extend_sys_msg_meta_dicts is not None:
sys_msg_meta_dicts = [{**sys_msg_meta_dict, **extend_sys_msg_meta_dict} for
sys_msg_meta_dict, extend_sys_msg_meta_dict in
zip(sys_msg_meta_dicts, extend_sys_msg_meta_dicts)]
self.assistant_sys_msg = SystemMessage(role_name=assistant_role_name, role_type=RoleType.DEFAULT,
meta_dict=sys_msg_meta_dicts[0],
content=assistant_role_prompt.format(**sys_msg_meta_dicts[0]))
self.user_sys_msg = SystemMessage(role_name=user_role_name, role_type=RoleType.DEFAULT,
meta_dict=sys_msg_meta_dicts[1],
content=user_role_prompt.format(**sys_msg_meta_dicts[1]))
self.assistant_agent: ChatAgent = ChatAgent(self.assistant_sys_msg, memory, model_type,
**(assistant_agent_kwargs or {}), )
self.user_agent: ChatAgent = ChatAgent(self.user_sys_msg,memory, model_type, **(user_agent_kwargs or {}), )
if with_critic_in_the_loop:
raise ValueError("with_critic_in_the_loop not available")
# if critic_role_name.lower() == "human":
# self.critic = Human(**(critic_kwargs or {}))
# else:
# critic_criteria = (critic_criteria or "improving the task performance")
# critic_msg_meta_dict = dict(critic_role=critic_role_name, criteria=critic_criteria,
# **sys_msg_meta_dicts[0])
# self.critic_sys_msg = sys_msg_generator.from_dict(critic_msg_meta_dict,
# role_tuple=(critic_role_name, RoleType.CRITIC), )
# self.critic = CriticAgent(self.critic_sys_msg, model_type, **(critic_kwargs or {}), )
else:
self.critic = None
def init_chat(self, phase_type: PhaseType = None,
placeholders=None, phase_prompt=None):
r"""Initializes the chat by resetting both the assistant and user
agents, and sending the system messages again to the agents using
chat messages. Returns the assistant's introductory message and the
user's response messages.
Returns:
A tuple containing an `AssistantChatMessage` representing the
assistant's introductory message, and a list of `ChatMessage`s
representing the user's response messages.
"""
if placeholders is None:
placeholders = {}
self.assistant_agent.reset()
self.user_agent.reset()
# refactored ChatDev
content = phase_prompt.format(
**({"assistant_role": self.assistant_agent.role_name} | placeholders)
)
retrieval_memory = self.assistant_agent.use_memory(content)
if retrieval_memory!= None:
placeholders["examples"] = retrieval_memory
user_msg = UserChatMessage(
role_name=self.user_sys_msg.role_name,
role="user",
content=content
# content here will be concatenated with assistant role prompt (because we mock user and send msg to assistant) in the ChatAgent.step
)
pseudo_msg = copy.deepcopy(user_msg)
pseudo_msg.role = "assistant"
self.user_agent.update_messages(pseudo_msg)
# here we concatenate to store the real message in the log
log_visualize(self.user_agent.role_name,
"**[Start Chat]**\n\n[" + self.assistant_agent.system_message.content + "]\n\n" + content)
return None, user_msg
def process_messages(
self,
messages: Sequence[ChatMessage],
) -> ChatMessage:
r"""Processes a list of chat messages, returning the processed message.
If multiple messages are provided and `with_critic_in_the_loop`
is `False`, raises a `ValueError`. If no messages are provided, also
raises a `ValueError`.
Args:
messages:
Returns:
A single `ChatMessage` representing the processed message.
"""
if len(messages) == 0:
raise ValueError("No messages to process.")
if len(messages) > 1 and not self.with_critic_in_the_loop:
raise ValueError("Got than one message to process. "
f"Num of messages: {len(messages)}.")
elif self.with_critic_in_the_loop and self.critic is not None:
processed_msg = self.critic.step(messages)
else:
processed_msg = messages[0]
return processed_msg
def step(
self,
user_msg: ChatMessage,
assistant_only: bool,
) -> Tuple[ChatAgentResponse, ChatAgentResponse]:
assert isinstance(user_msg, ChatMessage), print("broken user_msg: " + str(user_msg))
# print("assistant...")
user_msg_rst = user_msg.set_user_role_at_backend()
assistant_response = self.assistant_agent.step(user_msg_rst)
if assistant_response.terminated or assistant_response.msgs is None:
return (
ChatAgentResponse([assistant_response.msgs], assistant_response.terminated, assistant_response.info),
ChatAgentResponse([], False, {}))
assistant_msg = self.process_messages(assistant_response.msgs)
if self.assistant_agent.info:
return (ChatAgentResponse([assistant_msg], assistant_response.terminated, assistant_response.info),
ChatAgentResponse([], False, {}))
self.assistant_agent.update_messages(assistant_msg)
if assistant_only:
return (
ChatAgentResponse([assistant_msg], assistant_response.terminated, assistant_response.info),
ChatAgentResponse([], False, {})
)
# print("user...")
assistant_msg_rst = assistant_msg.set_user_role_at_backend()
user_response = self.user_agent.step(assistant_msg_rst)
if user_response.terminated or user_response.msgs is None:
return (ChatAgentResponse([assistant_msg], assistant_response.terminated, assistant_response.info),
ChatAgentResponse([user_response], user_response.terminated, user_response.info))
user_msg = self.process_messages(user_response.msgs)
if self.user_agent.info:
return (ChatAgentResponse([assistant_msg], assistant_response.terminated, assistant_response.info),
ChatAgentResponse([user_msg], user_response.terminated, user_response.info))
self.user_agent.update_messages(user_msg)
return (
ChatAgentResponse([assistant_msg], assistant_response.terminated, assistant_response.info),
ChatAgentResponse([user_msg], user_response.terminated, user_response.info),
)