RuntimeError: Code execution is set to be run in docker (default behaviour) but docker is not running.

[tyler-suard-parker]

Describe the issue

This error appears even when I have no plans to run any code. It is really annoying, and sometimes I can't set my environment variables to get rid of it.

Steps to reproduce

Create agents and run them

Screenshots and logs

No response

Additional Information

No response

[olgavrou]

Hi @tyler-suard-parker

This is a recent change that was made for safety around code execution, you can see how to disable the default behaviour here if you wish to do so.

If you are still having trouble can you please add some code so that we can see where the problem is, and let us know your dev operating system, etc

[sonichi]

Maybe we can add in the doc about disabling code execution entirely, by setting code_execution_config=False

[tyler-suard-parker]

Can we just set code execution to be disabled by default?

[olgavrou]

This would also be a breaking change; we could consider it for future releases. @sonichi I'll update the documentation

[austinhumes-valtech]

I just updated to the latest version 0.2.9 and received the error:

ERROR: Code execution is set to be run in docker (default behaviour) but docker is not running.
The options available are:
- Make sure docker is running (advised approach for code execution)
- Set "use_docker": False in code_execution_config
- Set AUTOGEN_USE_DOCKER to "0/False/no" in your environment variables

even though I had code_execution_config=False already, while not using Docker at all. Once I just started Docker Desktop and restarted my Autogen app, it ran without error. Does that mean that Docker just has to be running on the same machine, even if it's not used with Autogen?

Thanks in advance @olgavrou @sonichi

[olgavrou]

Hi @austinhumes-valtech
If you had code_execution_config set to False then it shouldn't complain about docker. Could you provide a sample of your code maybe so that we can investigate?

[austinhumes-valtech]

Hi @austinhumes-valtech If you had code_execution_config set to False then it shouldn't complain about docker. Could you provide a sample of your code maybe so that we can investigate?

Hi @olgavrou thanks for the quick response! I'm using a group chat. Here's some code samples:

class AutogenChat():
    def __init__(self, chat_id=None, websocket=None):
        self.websocket = websocket
        self.chat_id = chat_id
        self.client_sent_queue = asyncio.Queue()
        self.client_receive_queue = asyncio.Queue()  
        
        self.user_proxy = UserProxyWebAgent( 
            name="user_proxy",
            human_input_mode="ALWAYS", 
            system_message="""
                Agent to represent the human user.
                Always provide the user with a easy to understand response, rather than just data. For example, if a large amount of data is returned from a function call, provide the user with a summary of the results instead of the raw API response, for example: 
                1. Item 1
                2. Item 2
                3. Item 3
            """,
            max_consecutive_auto_reply=5,
            is_termination_msg=lambda x: x.get("content", "") and x.get("content", "").rstrip().endswith("TERMINATE"),
            code_execution_config=False
        )

   ...

class UserProxyWebAgent(autogen.UserProxyAgent):
    def __init__(self, *args, **kwargs):
        super(UserProxyWebAgent, self).__init__(*args, **kwargs)
        self._reply_func_list = []
        self.register_reply([Agent, None], ConversableAgent.generate_oai_reply)
        self.register_reply([Agent, None], ConversableAgent.generate_code_execution_reply)
        self.register_reply([Agent, None], ConversableAgent.generate_function_call_reply)
        self.register_reply([Agent, None], UserProxyWebAgent.a_check_termination_and_human_reply)

    ...

Hope that helps! Thanks!

[ekzhu]

It only shows the set up for the user proxy, what about the assistant agent? Did you set the code_execution_config=False for the assistant agent?

Also, I see you are re-registering the generate_code_execution_reply function. You probably want to remove that and only register it depending on whether you are turning on code execution.

[sonichi]

@austinhumes-valtech does this error still occur with the latest version? If so, could you share a little reproducible example?

[doursand]

I managed to solved this issue adding a "use_docker": False inside the code_execution_config dict like this :

user_proxy = autogen.UserProxyAgent( name="user_proxy", human_input_mode="NEVER", max_consecutive_auto_reply=10, is_termination_msg=lambda x: x.get("content", "").rstrip().endswith("TERMINATE"), code_execution_config={"work_dir": "web", "use_docker": False}, llm_config=llm_config_mistral, system_message="""Reply TERMINATE if the task has been solved at full satisfaction. Otherwise, reply CONTINUE, or the reason why the task is not solved yet.""" )

[olgavrou]

Great! this is the intended workaround :) I'll close the issue but feel free to reopen if another issue occurs

[g-o-o-g-l-e]

I got RuntimeError: Please ensure Docker is running on your system.

[g-o-o-g-l-e]

on the lastest version

[g-o-o-g-l-e]

my code is from docker_games import Minecraft

minecraft = Minecraft('creepers_go_boom', '2GB', '/home/stefano/mc_data')
minecraft.add_ports(25565, 25565)
minecraft.accept_license()
minecraft.online_mode()

minecraft.start()

[ekzhu]

@g-o-o-g-l-e you need to install docker on your machine

[honestAnt]

This is because the docker client integrated with autogen uses the default configuration (mine is macos, the default port is 2375), and DOCKER_HOST needs to be modified to solve the problem.

So you can add DOCK_HOST configuration to the environment variable to solve the problem:

[g-o-o-g-l-e]

thanks

[AbhijitManepatil]

I managed to solved this issue adding a "use_docker": False inside the code_execution_config dict like this :

user_proxy = autogen.UserProxyAgent( name="user_proxy", human_input_mode="NEVER", max_consecutive_auto_reply=10, is_termination_msg=lambda x: x.get("content", "").rstrip().endswith("TERMINATE"), code_execution_config={"work_dir": "web", "use_docker": False}, llm_config=llm_config_mistral, system_message="""Reply TERMINATE if the task has been solved at full satisfaction. Otherwise, reply CONTINUE, or the reason why the task is not solved yet.""" )

"use_docker": False works