Motivation

Now CAMEL apps only support chatting between two autonomous agents. It is important to support more chat modes for different use cases.

• Autonomous chat with more than two agents
• Chat with Human-in-the-loop (#77)
• Support chatting with Human-in-the-loop in UI

Thanks @marchowardbegins for the discussion #57 (comment)

Would you guys like them?

https://github.com/lightaime/camel/blob/ba3ec7603ebc735600e32379df235e627c4141c8/camel/agents/chat_agent.py#L93

The prompt in the project contains<YOU_ SOLUTION>information, but no alternative logic was found in the code.

Hello, I noticed that you have Gradio for the demo however can we also get this for local use?

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

It would be helpful for people who want to try the demo out for coding tasks

Solution

No response

Alternatives

No response

Additional context

No response

I got the single_agent.py to work. No problems using conda Python 3.10.10. or another Python 3.10.7 environment But I did not get further. I get the following error trying out e.g. role_playing_multiprocess.py:
role_playing_multiprocess.py, line 162, in main
array_idx = int(os.environ.get('SLURM_ARRAY_TASK_ID'))
TypeError: int() argument must be a string, a bytes-like object or a real number, not 'NoneType'
Am I missing something?

Add a dummy class to mock OpenAI API so the OpenAI API is not used during the tests to make them faster and cheaper.

I cant run the script , mentioning that i'm a newbie and please help
iloading@DESKTOP-CPIUK2P:/mnt/c/Users/user/Desktop/camel-master/camel-master/examples/ai_society$ python3 role_playing.py
Traceback (most recent call last):
File "/mnt/c/Users/user/Desktop/camel-master/camel-master/examples/ai_society/role_playing.py", line 3, in
from camel.agents import RolePlaying
ModuleNotFoundError: No module named 'camel'

Required prerequisites

• I have read the documentation https://camel-ai.github.io/camel/camel.html.
• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

What version of camel are you using?

master/HEAD

System information

master/HEAD

Problem description

Running mypy on the code highlights around 30 bugs in the code found my linter.

Reproducible example code

mypy --namespace-packages -p camel
mypy --namespace-packages -p test
mypy --namespace-packages -p examples
mypy --namespace-packages -p apps

Traceback

No response

Expected behavior

No response

Additional context

No response

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

The current data schema is implemented with data classes that do not enforce typing checking and data validation at runtime. We can introduce better data validation and settings management by replacing data classes with pydantic.

Solution

Use pydantic dataclasses: https://docs.pydantic.dev/latest/usage/dataclasses/

Make all dataclasses in the project pydantic dataclasses, frozen and ordered.

@dataclass(frozen=True, order=True)

My need extra kwonly=True.

Alternatives

No response

Additional context

No response

Required prerequisites

• I have read the documentation https://camel-ai.github.io/camel/camel.html.
• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Questions

According to your design , wo need to choose the user role and assistantt role first. I think can we not need think what role wo need, the camel can autonomously choose the roles according to the preliminary idea?

Would love to be able to use local LLMS like Alpaca and Llama.

Just had a try at http://agents.camel-ai.org/

But what the model spits out contains <br> html tag which makes them a lot harder to use

Thank! Your work is amazing, it opens up a lot of possibilities, especially in eduacting onself in learning something new w/o knowing much but just a couples ideas.
This work deserve more attention!

Required prerequisites

• I have read the documentation https://camel-ai.github.io/camel/camel.html.
• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

What version of camel are you using?

master/HEAD

System information

Any

Problem description

Human agent uses input() as core functionality which breaks the pureness of the package. All interaction must go via the API.

Reproducible example code

Traceback

No response

Expected behavior

No response

Additional context

No response

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

Now the code execution of Python is using bare exec, which can cause arbitrary code execution and system crash. Moreover, the value in exec cannot be captured by outer program, which add difficulty for some agent implementation.

Solution

We can use ast in python to analysis generated code and control the execution of interpreter.

Alternatives

No response

Additional context

No response

Hello,
The demo that is set up at https://f04f43d534b998d61a.gradio.live/ is currently down.
It only displays the following message: No interface is running right now

python examples/ai_society/role_playing.py
Traceback (most recent call last):
File "examples/ai_society/role_playing.py", line 3, in
from camel.agents import RolePlaying
ModuleNotFoundError: No module named 'camel'

The current role-playing session usually ends due to 'max_tokens_exceeded'. To complete more complex tasks, external memory is needed.

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

The current EmbodiedAgent is not yet integrated into RolePlaying with other agents yets.

Solution

No response

Alternatives

No response

Additional context

No response

Hey I'm trying to edit the default assistant/user prompts in camel\agents\prompts\ai_society but I'm getting the error in the attached image:

File "C:\tools\camel\camel\generators.py", line 46, in validate_meta_dict_keys
raise ValueError("The keys of the meta_dict should be in "
ValueError: The keys of the meta_dict should be in set(). Got {'<USER_ROLE>', '', '<ASSISTANT_ROLE>'} instead.

Here are the custom Assistant/User prompt I want to include:
Proteus 4.2 - OmniCompetent Assistant: https://flowgpt.com/prompt/Ku_FbNzQsd_itThT7xhY6
Tech writer/Super AI researcher - Dr. Ada Turing: https://flowgpt.com/prompt/3EB4JHHhJPiP51f9fA9Bf

The PersRubric/Omnicomps in the above are quite interesting and...effective, I want to use Camel to test how effective they actually are by doing a few hundred runs with variations of the above vs default settings.

Subtasks:

1. Make RolePlaying.step() return a single message (user or assistant separately). If critic is enabled, return alternatives to critize separately as single messages.
2. Internally maintain a state machine that tracks if Human Critic is awaited, and throw an Exception if step() is called without submitting the response by Human Critic
3. add two API methods

isResponseAwaited() -> bool
submitResponse(str) -> None

Motivation

The ability to call tools would enhance CAMEL's ability and be handy for developers. Here are some tools we prioritize to integrate. Some of them are suggested by GPT-4:

• Search tools: Google API (#274), Wikipedia: (#213)
• Python interpreter (#182)
• HuggingFace Agent (#123)
• Function Call (#187)
• Weather API (@yiyiyi0817, Weather API - OpenWeatherMap)
• Github
• WolframAlpha
• UnstructuredIO
• Google Map
• Google Drive API
• Twitter API
• Discord API
• WeChat API
• WhatsApp API
• Facebook Graph API
• Youtube Data api
• Zoom API
• Spotify API
• Twilio API
• Slack API
• Shopify API
• Resend
• SendGrid API
• Reddit API
• Airtable API
• Yelp API
• TripAdvisor API
• Skyscanner API
• Booking.com API
• Text-to-image
• Text-to-speech
• LangChain Agent
• Duckduck go search
• Gorilla Agent (#146)
• Terminal
• Docker
• OpenCV

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

Now RolePlaying makes a hard stop if max token length is reached. The else branch of

if num_tokens < self.model_token_limit:

is problematic for users. message_window_size is not a good solution since it makes a user guess how many messages may fit into the context and they may not fit anyway and ruin the roleplay.

Solution

We can give user the choice of:

1. Simple-compress the chat by keeping the system prompt and discarding the oldest messages in the chat. Must guarantee fitting into model_token_limit. If needed must truncate the user and assistant messages by dropping the middle or the tail of a message. If needed it must even truncate the system message in the same manner.
2. Semantically compress the chat by summarizing messages with extra chatgpt calls (later).
3. Exit the chat as we do now

message_window_size fuctionality must be removed. Hard stop must be not be issued for option 1.

Alternatives

No response

Additional context

No response

Required prerequisites

• I have read the documentation https://camel-ai.github.io/camel/camel.html.
• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

What version of camel are you using?

0.1.0

System information

OS: Windows

Problem description

Users attempting to use the camel project on Windows have reported issues with setting the OpenAI API key as an environment variable. The current instructions in the README seem to be more suited for Unix-like environments (like Linux or macOS), causing confusion among Windows users.

Reproducible example code

1. Clone and install the camel project as per the README.
2. Attempt to set the OpenAI API key as an environment variable using the export command in Windows PowerShell or Command Prompt.
3. The export command is not recognized, and the environment variable is not set.
4. As a result, running scripts that require the OpenAI API key (like examples/ai_society/role_playing.py) results in errors.

Traceback

The `export` command is not recognized in Windows PowerShell or Command Prompt. Users have to use a different command (`$env:OPENAI_API_KEY = "your-api-key"`) to set the environment variable. This is not currently mentioned in the README.

Expected behavior

Windows users should be able to set the OpenAI API key as an environment variable in their terminal, allowing scripts to access the key as needed.

Additional context

Update the README and any other relevant documentation to include specific instructions for setting environment variables on Windows. This should cover both Command Prompt and PowerShell syntax.

Motivation

We need to clean up code so that the API is easier to study and use. List of required changes:

• Remove UserChatMessage and AssistantChatMessage
• Remove AssistantSystemMessage and UserSystemMessage
• Rename role from ChatMessage into role_at_backend
• Remove role from ChatMessage altogether
• Make BaseMessage, ChatMessage, SystemMessage a frozen=True dataclass
• Remove def __getattribute__ from BaseMessage (#168)
• Replace type of content in BaseMessage with TextPrompt
• Rename model to model_type
• Fix mode_type typos (#166, #169)
• Extract RolePlaying into a separate societies package
• Strict signature match for BaseAgent child classes

Solution

No response

Alternatives

No response

Additional context

No response

• Turning on and off
• Allow user to tune word_limit

Followed instructions to try and test the roleplaying example but it just errors out. would appreciate some guidance. Do we need to change or move the examples out of the folder - the instructions appear to suggest you just run the example demo from its location. GPT 4 suggestions were a bust.

PS D:\camel\camel\agent> python .\role_playing.py
Traceback (most recent call last):
File "D:\camel\camel\agent\role_playing.py", line 7, in
from .chat_agent import ChatAgent
ImportError: attempted relative import with no known parent package
PS D:\camel\camel\agent>

I also would like the addition of a free, open source LLM model locally for Camel.

——-

EDIT May 26, 2023

Perhaps this is an even better alternative than HuggingChat :

Gorilla - LLM with massive connected API !!!!!

https://gorilla.cs.berkeley.edu/

https://twitter.com/intuitmachine/status/1661812484062281744?s=46&t=X6lS_Zp1k_p2yWVr44f_7A

——-

My suggestion:

Perhaps HuggingChat v0.2 is a good alternative:

https://huggingface.co/chat/

HuggingChat v0.2Making the community's best AI chat models available to everyone.

NEWChat UI is now open sourced on GitHubGitHub repohttps://github.com/huggingface/chat-ui

Current Model
OpenAssistant/oasst-sft-6-llama-30bModelhttps://huggingface.co/OpenAssistant/oasst-sft-6-llama-30b-xor

Dataset
https://huggingface.co/datasets/OpenAssistant/oasst1

Website
https://open-assistant.io/

But because things are changing rapidly in these day and age, perhaps there are already other alternatives !
My application would be to use free open source LLMs instead of the OpenAI LLM that in the practical example that is being showcased here:

Camel + LangChain for Synthetic Data & Market Research
https://www.youtube.com/watch?v=GldMMK6-_-g

COLAB Notebook:
https://colab.research.google.com/drive/1BuudlvBrKBl1bNhMp-aB5uOW62-JlrSY?usp=sharing

Or in the example on your website:
camel_demo.ipynb
https://colab.research.google.com/drive/1AzP33O8rnMW__7ocWJhVBXjKziJXPtim?usp=sharing

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

The current EmbodiedAgent and HuggingFaceToolAgent just realize a very basic functionality. More features can be added to extend the action space.

Solution

No response

Alternatives

No response

Additional context

No response

Required prerequisites

• I have read the documentation https://camel-ai.github.io/camel/camel.html.
• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

What version of camel are you using?

master/HEAD

System information

Problem description

examples\ai_society\role_playing_with_critic.py

Reproducible example code

The Python snippets:

Command lines:

python examples\ai_society\role_playing_with_critic.py

Traceback

Exception has occurred: AttributeError       (note: full exception trace is shown but execution is paused at: __getattribute__)
'ChatMessage' object has no attribute '__deepcopy__'
  File "C:\Users\dmitrii\_KAUST\git\camel\camel\messages\base.py", line 105, in __getattribute__ (Current frame)
    return super().__getattribute__(name)
  File "C:\Users\dmitrii\_KAUST\git\camel\camel\agents\critic_agent.py", line 168, in step
    input_msg = copy.deepcopy(meta_chat_message)
  File "C:\Users\dmitrii\_KAUST\git\camel\camel\agents\role_playing.py", line 238, in process_messages
    processed_msg = self.critic.step(messages)
  File "C:\Users\dmitrii\_KAUST\git\camel\camel\agents\role_playing.py", line 274, in step
    user_msg = self.process_messages(user_response.msgs)
  File "C:\Users\dmitrii\_KAUST\git\camel\examples\ai_society\role_playing_with_critic.py", line 57, in main
    assistant_response, user_response = role_play_session.step(
  File "C:\Users\dmitrii\_KAUST\git\camel\examples\ai_society\role_playing_with_critic.py", line 84, in <module>
    main()
AttributeError: 'ChatMessage' object has no attribute '__deepcopy__'

Expected behavior

No response

Additional context

No response

Traceback (most recent call last):
  File "/Users/jonchui/code/chatgpt/camel/camelMe.py", line 37, in <module>
    main(args.prompt, args.assistant, args.user)
  File "/Users/jonchui/code/chatgpt/camel/camelMe.py", line 24, in main
    (assistant_msg, _, _), (user_msg, _, _) = role_play_session.step(assistant_msg)
  File "/Users/jonchui/code/chatgpt/camel/camel/agents/role_playing.py", line 113, in step
    raise RuntimeError("User agent is terminated.")
RuntimeError: User agent is terminated.

I'm going to see if CAMEL can fix itself ;)

#!/usr/bin/env python

import os

import openai

openai.api_key = "[REDACTED]"

import time

from colorama import Fore


def print_text_animated(text):
    for char in text:
        print(char, end="", flush=True)
        #time.sleep(0.02)

from camel.agents import RolePlaying

task_prompt = '''
DEBUG and fix this error that keeps happening when I try to run the code:

RuntimeError("User agent is terminated.")

Traceback (most recent call last):
File "/Users/jonchui/code/chatgpt/camel/camelMe.py", line 37, in
main(args.prompt, args.assistant, args.user)
File "/Users/jonchui/code/chatgpt/camel/camelMe.py", line 24, in main
(assistant_msg, _, _), (user_msg, _, _) = role_play_session.step(assistant_msg)
File "/Users/jonchui/code/chatgpt/camel/camel/agents/role_playing.py", line 113, in step
raise RuntimeError("User agent is terminated.")
RuntimeError: User agent is terminated.

last function is 

def step(
self,
assistant_msg: ChatMessage,
) -> Tuple[Tuple[ChatMessage, bool, Dict], Tuple[ChatMessage, bool, Dict]]:
user_msgs, user_terminated, user_info = self.user_agent.step(
assistant_msg)
if user_terminated:
raise RuntimeError("User agent is terminated.")
user_msg = user_msgs[0]
user_msg.role = "user"

    (assistant_msgs, assistant_terminated,
     assistant_info) = self.assistant_agent.step(user_msg)
    if assistant_terminated:
        raise RuntimeError("Assistant agent is terminated.")
    assistant_msg = assistant_msgs[0]
    assistant_msg.role = "user"

    return (
        (assistant_msg, assistant_terminated, assistant_info),
        (user_msg, user_terminated, user_info),
    )

the github project is here:
https://github.com/lightaime/camel

'''
print(Fore.YELLOW + f"Original task prompt:\n{task_prompt}\n")
role_play_session = RolePlaying("Computer Programmer", "Software Engineer", task_prompt)
print(Fore.CYAN + f"Specified task prompt:\n{role_play_session.task_prompt}\n")

chat_turn_limit, n = 1000, 0
assistant_msg, _ = role_play_session.init_chat()
while n < chat_turn_limit:
    n += 1
    (assistant_msg, _, _), (user_msg, _, _) = role_play_session.step(assistant_msg)
    print_text_animated(Fore.BLUE + f"AI User:\n\n{user_msg.content}\n\n")
    print_text_animated(Fore.GREEN + f"AI Assistant:\n\n{assistant_msg.content}\n\n")
    if "<CAMEL_TASK_DONE>" in user_msg.content:
        break

It is important for the users to learn the basic concepts and the core modules of CAMEL. I made a markdown tutorial for TextPrompt with the help of ChatGPT. Here is how I prompt it:

Make a markdown tutorial for the `TextPrompt` class based on its pytest functions:

{the `TextPrompt` class}

{its pytest functions}

ChatGPT generated a draft for the tutorial then I did some cleanup and editing. Although it is not perfect, it greatly accelerated the process. Almost every core modules already have pytest functions implemented. We can use a similar approach to generate tutorials for other modules. Let me know if you want to help out.

Note that: the pytest functions can be found here, after we create the markdown file (*.md), we can put it under docs/get_started and add it to docs/index.rst. Make sure to test if it works as expected before you make it pull request following the instructions of Building Documentation.

Here are some important modules we can work on:

• Overview of CAMEL repo design
• camel/prompts/base.py::TextPrompt
• camel/messages.py
• camel/agents/chat_agent.py (Memory, Tool usage, RAG, Code Interpreter)
• camel/agents/task_agent.py
• camel/agents/role_playing.py (Memory, Tool usage, RAG, Code Interpreter)
• camel/agents/critic_agent.py
• camel/generators.py::SystemMessageGenerator
• camel/prompts/ai_society.py::AISocietyPromptTemplateDict

Thanks!

Fix temporary solutions in Code example #32.

• Remove CodeChatAgent (#49)
• Remove CodeSystemMessageGenerator (#49)
• Remove CodeTaskPromptGenerator

(camel) C:\Users\joshu\Documents\_aipython\_useful\CAMEL agnet\camel\camel\agents>python role_playing.py
Traceback (most recent call last):
  File "C:\Users\joshu\Documents\_aipython\_useful\CAMEL agnet\camel\camel\agents\role_playing.py", line 8, in <module>
    from .chat_agent import ChatAgent
ImportError: attempted relative import with no known parent package

any help would be good.

Required prerequisites

• I have read the documentation https://camel-ai.github.io/camel/camel.html.
• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

What version of camel are you using?

0.1.0

System information

3.7.3 (default, Mar 27 2019, 22:11:17)
[GCC 7.3.0] linux

Problem description

Task files could be downloaded automatically if no exist for examples/ai_society/role_playing_multiprocess.py and examples/code/role_playing_multiprocess.py

Reproducible example code

The Python snippets:

python examples/ai_society/role_playing_multiprocess.py
python examples/code/role_playing_multiprocess.py

Traceback

No response

Expected behavior

No response

Additional context

No response

Required prerequisites

• I have read the documentation https://camel-ai.github.io/camel/camel.html.
• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

What version of camel are you using?

latest

System information

Google Colab notebook

Problem description

Colab notebook fails to run adding "from colorama import Fore" fixes. Thanks.

Reproducible example code

The Python snippets:

Command lines:

Extra dependencies:

Steps to reproduce:

Traceback

No response

Expected behavior

No response

Additional context

No response

thread '' panicked at 'assertion failed: encoder.len() == decoder.len()', src\lib.rs:458:9
note: run with RUST_BACKTRACE=1 environment variable to display a backtrace

File ".\camel\agents\chat_agent.py", line 71, in step
num_tokens = num_tokens_from_messages(openai_messages, self.model)
encoding = tiktoken.encoding_for_model(model.value)
return get_encoding(encoding_name)
enc = Encoding(**constructor())
self._core_bpe = _tiktoken.CoreBPE(mergeable_ranks, special_tokens, pat_str)

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

Examples for generating math and science datasets are missing

Solution

No response

Alternatives

No response

Additional context

No response