You can also check this cookbook in colab here

Star us on Github, join our Discord or follow our X

1. Concept

The society module is one of the core modules of CAMEL. By simulating the information exchange process, this module studies the social behaviors among agents.

Currently, the society module aims to enable agents to collaborate autonomously toward completing tasks while keeping consistent with human intentions and requiring minimal human intervention. It includes two frameworks: RolePlaying and BabyAGI, which are used to run the interaction behaviors of agents to achieve objectives.

Taking RolePlaying as an example, this framework was designed in an instruction-following manner. These roles independently undertake the responsibilities of executing and planning tasks, respectively. The dialogue is continuously advanced through a turn-taking mechanism, thereby collaborating to complete tasks. The main concepts include:

  • Task: a task can be as simple as an idea, initialized by an inception prompt.

  • AI User: the agent who is expected to provide instructions.

  • AI Assistant: the agent who is expected to respond with solutions that fulfill the instructions.

2. Types

2.1 RolePlaying

RolePlaying is a unique cooperative agent framework of CAMEL. Through this framework, agents in CAMEL overcome numerous challenges, such as role flipping, assistant repeats instructions, flake replies, infinite loop of messages, and conversation termination conditions.

When using RolePlaying framework in CAMEL, predefined prompts are used to create unique initial settings for different agents. For example, if the user wants to initialize an assistant agent, the agent will be initialized with the following prompt.

  • Never forget you are a ASSISTANT_ROLE and I am a USER_ROLE.

    This assigns the chosen role to the assistant agent and provides it with information about the user’s role.

  • Never flip roles! Never instruct me!

    This prevents agents from flipping roles. In some cases, we have observed the assistant and the user switching roles, where the assistant suddenly takes control and instructs the user, and the user follows those instructions.

  • You must decline my instruction honestly if you cannot perform the instruction due to physical, moral, legal reasons or your capability and explain the reasons.

    This prohibits the agent from producing harmful, false, illegal, and misleading information.

  • Unless I say the task is completed, you should always start with: Solution: <YOUR_SOLUTION>. <YOUR_SOLUTION> should be specific, and provide preferable implementations and examples for task-solving.

    This encourages the assistant to always responds in a consistent format, avoiding any deviation from the structure of the conversation, and preventing vague or incomplete responses, which we refer to as flake responses, such as “I will do something”.

  • Always end your solution with: Next request.

    This ensures that the assistant keeps the conversation going by requesting a new instruction to solve.

RolePlaying Attributes

AttributeTypeDescription
assistant_role_namestrThe name of the role played by the assistant.
user_role_namestrThe name of the role played by the user.
critic_role_namestrThe name of the role played by the critic.
task_promptstrA prompt for the task to be performed.
with_task_specifyboolWhether to use a task specify agent.
with_task_plannerboolWhether to use a task planner agent.
with_critic_in_the_loopboolWhether to include a critic in the loop.
critic_criteriastrCritic criteria for the critic agent.
modelBaseModelBackendThe model backend to use for generating responses.
task_typeTaskTypeThe type of task to perform.
assistant_agent_kwargsDictAdditional arguments to pass to the assistant agent.
user_agent_kwargsDictAdditional arguments to pass to the user agent.
task_specify_agent_kwargsDictAdditional arguments to pass to the task specify agent.
task_planner_agent_kwargsDictAdditional arguments to pass to the task planner agent.
critic_kwargsDictAdditional arguments to pass to the critic.
sys_msg_generator_kwargsDictAdditional arguments to pass to the system message generator.
extend_sys_msg_meta_dictsList[Dict]A list of dicts to extend the system message meta dicts with.
extend_task_specify_meta_dictDictA dict to extend the task specify meta dict with.
output_languagestrThe language to be output by the agents.

2.2 BabyAGI

Babyagi is framework from “Task-driven Autonomous Agent

3. Get Started

Installation

Ensure you have CAMEL AI installed in your Python environment:

!pip install "camel-ai==0.2.16"

Setting Up API Keys

You’ll need to set up your API keys for OpenAI.

import os
from getpass import getpass

# Prompt for the API key securely
openai_api_key = getpass('Enter your API key: ')
os.environ["OPENAI_API_KEY"] = openai_api_key

Alternatively, if running on Colab, you could save your API keys and tokens as Colab Secrets, and use them across notebooks.

To do so, comment out the above manual API key prompt code block(s), and uncomment the following codeblock.

⚠️ Don’t forget granting access to the API key you would be using to the current notebook.

# import os
# from google.colab import userdata

# os.environ["OPENAI_API_KEY"] = userdata.get("OPENAI_API_KEY")

3.1. Using RolePlaying

from colorama import Fore

from camel.societies import RolePlaying
from camel.utils import print_text_animated
def main(model=None, chat_turn_limit=10) -> None:
# Initial the role-playing session on developing a trading bot task with default model (`GPT_4O_MINI`)
    task_prompt = "Develop a trading bot for the stock market"
    role_play_session = RolePlaying(
        assistant_role_name="Python Programmer",
        assistant_agent_kwargs=dict(model=model),
        user_role_name="Stock Trader",
        user_agent_kwargs=dict(model=model),
        task_prompt=task_prompt,
        with_task_specify=True,
        task_specify_agent_kwargs=dict(model=model),
    )

# Output initial message with different colors.
    print(
        Fore.GREEN
        + f"AI Assistant sys message:\n{role_play_session.assistant_sys_msg}\n"
    )
    print(
        Fore.BLUE + f"AI User sys message:\n{role_play_session.user_sys_msg}\n"
    )

    print(Fore.YELLOW + f"Original task prompt:\n{task_prompt}\n")
    print(
        Fore.CYAN
        + "Specified task prompt:"
        + f"\n{role_play_session.specified_task_prompt}\n"
    )
    print(Fore.RED + f"Final task prompt:\n{role_play_session.task_prompt}\n")

    n = 0
    input_msg = role_play_session.init_chat()

# Output response step by step with different colors.
# Keep output until detect the terminate content or reach the loop limit.
    while n < chat_turn_limit:
        n += 1
        assistant_response, user_response = role_play_session.step(input_msg)

        if assistant_response.terminated:
            print(
                Fore.GREEN
                + (
                    "AI Assistant terminated. Reason: "
                    f"{assistant_response.info['termination_reasons']}."
                )
            )
            break
        if user_response.terminated:
            print(
                Fore.GREEN
                + (
                    "AI User terminated. "
                    f"Reason: {user_response.info['termination_reasons']}."
                )
            )
            break

        print_text_animated(
            Fore.BLUE + f"AI User:\n\n{user_response.msg.content}\n"
        )

        if "CAMEL_TASK_DONE" in user_response.msg.content:
            break

        print_text_animated(
            Fore.GREEN + "AI Assistant:\n\n"
            f"{assistant_response.msg.content}\n"
        )



        input_msg = assistant_response.msg

if __name__ == "__main__":
    main()