Skip to content

Instantly share code, notes, and snippets.

@mberman84

mberman84/gist:c95d69263d9c2ceb7d56cf336f13ae02

Created October 26, 2023 19:03
Show Gist options
  • Save mberman84/c95d69263d9c2ceb7d56cf336f13ae02 to your computer and use it in GitHub Desktop.
Save mberman84/c95d69263d9c2ceb7d56cf336f13ae02 to your computer and use it in GitHub Desktop.
Download ZIP
AutoGen + MemGPT Code
Raw
gistfile1.txt
import os
import autogen
import memgpt.autogen.memgpt_agent as memgpt_autogen
import memgpt.autogen.interface as autogen_interface
import memgpt.agent as agent
import memgpt.system as system
import memgpt.utils as utils
import memgpt.presets as presets
import memgpt.constants as constants
import memgpt.personas.personas as personas
import memgpt.humans.humans as humans
from memgpt.persistence_manager import InMemoryStateManager, InMemoryStateManagerWithPreloadedArchivalMemory, InMemoryStateManagerWithFaiss
import openai
openai.api_key = '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'
config_list = [
{
'model': 'gpt-4'
},
]
llm_config = {"config_list": config_list, "seed": 42}
user_proxy = autogen.UserProxyAgent(
name="User_proxy",
system_message="A human admin.",
code_execution_config={"last_n_messages": 2, "work_dir": "groupchat"},
)
interface = autogen_interface.AutoGenInterface() # how MemGPT talks to AutoGen
persistence_manager = InMemoryStateManager()
persona = "I\'m a 10x engineer at a FAANG tech company."
human = "I\'m a team manager at a FAANG tech company."
memgpt_agent = presets.use_preset(presets.DEFAULT, 'gpt-4', persona, human, interface, persistence_manager)
# MemGPT coder
coder = memgpt_autogen.MemGPTAgent(
name="MemGPT_coder",
agent=memgpt_agent,
)
# non-MemGPT PM
pm = autogen.AssistantAgent(
name="Product_manager",
system_message="Creative in software product ideas.",
llm_config=llm_config,
)
groupchat = autogen.GroupChat(agents=[user_proxy, coder, pm], messages=[], max_round=12)
manager = autogen.GroupChatManager(groupchat=groupchat, llm_config=llm_config)
user_proxy.initiate_chat(manager, message="First send the message 'Let's go Mario!'")
@hijkzzz
Copy link

hijkzzz commented Nov 17, 2023

I met the problem AttributeError: 'dict' object has no attribute 'model_endpoint_type'

@slaughters85j
Copy link

slaughters85j commented Nov 20, 2023 •
edited

I found success after mixing a few of the suggestions above.

  1. per Matt's video, I created a conda environment using 3.11.3
  2. I used the absolute path to the conda python env via 'which python' to install only pymemgpt and pyautogen
  3. Modified the import of presets to
import memgpt.presets.presets as presets
  1. Modified the code block in app.py to
interface = autogen_interface.AutoGenInterface() # how MemGPT talks to AutoGen
persistence_manager = InMemoryStateManager()
persona = "I am an engineer at Att3st tech company."
human = "I am a team manager at Att3st tech company."
agent_config = llm_config
memgpt_agent = presets.use_preset(presets.DEFAULT_PRESET, model='gpt-4', persona=persona, human=human, interface=interface, persistence_manager=persistence_manager,agent_config=agent_config)
  1. Modified the MemGPTAgent (memgpt_agent.py) file start at line 229 to:
    async def _generate_reply_for_user_message(
        self,
        messages: Optional[List[Dict]] = None,
        sender: Optional[Agent] = None,
        config: Optional[Any] = None,
    ) -> Tuple[bool, Union[str, Dict, None]]:
        self.agent.interface.reset_message_list()

        new_messages = self.find_new_messages(messages)
        if len(new_messages) > 1:
            if self.concat_other_agent_messages:
                # Combine all the other messages into one message
                user_message = "\n".join([self.format_other_agent_message(m) for m in new_messages])
            else:
                # Extend the MemGPT message list with multiple 'user' messages, then push the last one with agent.step()
                self.agent.messages.extend(new_messages[:-1])
                user_message = new_messages[-1]
        elif len(new_messages) == 1:
            user_message = new_messages[0]
        else:
            return True, self._default_auto_reply

        # Package the user message
        user_message = system.package_user_message(user_message)

        # Send a single message into MemGPT
        while True:
            (
                new_messages,
                heartbeat_request,
                function_failed,
                token_warning,
            ) = await self.agent.step(user_message, first_message=False, skip_verify=self.skip_verify)

It's now working.

@Geetha-muffin
Copy link

Geetha-muffin commented Nov 24, 2023 •
edited

I found success after mixing a few of the suggestions above.

  1. per Matt's video, I created a conda environment using 3.11.3
  2. I used the absolute path to the conda python env via 'which python' to install only pymemgpt and pyautogen
  3. Modified the import of presets to
import memgpt.presets.presets as presets
  1. Modified the code block in app.py to
interface = autogen_interface.AutoGenInterface() # how MemGPT talks to AutoGen
persistence_manager = InMemoryStateManager()
persona = "I am an engineer at Att3st tech company."
human = "I am a team manager at Att3st tech company."
agent_config = llm_config
memgpt_agent = presets.use_preset(presets.DEFAULT_PRESET, model='gpt-4', persona=persona, human=human, interface=interface, persistence_manager=persistence_manager,agent_config=agent_config)
  1. Modified the MemGPTAgent (memgpt_agent.py) file start at line 229 to:
    async def _generate_reply_for_user_message(
        self,
        messages: Optional[List[Dict]] = None,
        sender: Optional[Agent] = None,
        config: Optional[Any] = None,
    ) -> Tuple[bool, Union[str, Dict, None]]:
        self.agent.interface.reset_message_list()

        new_messages = self.find_new_messages(messages)
        if len(new_messages) > 1:
            if self.concat_other_agent_messages:
                # Combine all the other messages into one message
                user_message = "\n".join([self.format_other_agent_message(m) for m in new_messages])
            else:
                # Extend the MemGPT message list with multiple 'user' messages, then push the last one with agent.step()
                self.agent.messages.extend(new_messages[:-1])
                user_message = new_messages[-1]
        elif len(new_messages) == 1:
            user_message = new_messages[0]
        else:
            return True, self._default_auto_reply

        # Package the user message
        user_message = system.package_user_message(user_message)

        # Send a single message into MemGPT
        while True:
            (
                new_messages,
                heartbeat_request,
                function_failed,
                token_warning,
            ) = await self.agent.step(user_message, first_message=False, skip_verify=self.skip_verify)

It's now working.

@slaughters85j Could you please share the complete code with requirements file ?

@Mugunthapranav
Copy link

User_proxy (to chat_manager):

First send the message 'Let's go Mario!'

Traceback (most recent call last):
File "C:\Users\Muguntha Pranav M\PycharmProjects\Auto_mem\main.py", line 51, in
user_proxy.initiate_chat(manager, message="First send the message 'Let's go Mario!'")
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\autogen\agentchat\conversable_agent.py", line 531, in initiate_chat
self.send(self.generate_init_message(**context), recipient, silent=silent)
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\autogen\agentchat\conversable_agent.py", line 334, in send
recipient.receive(message, self, request_reply, silent)
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\autogen\agentchat\conversable_agent.py", line 462, in receive
reply = self.generate_reply(messages=self.chat_messages[sender], sender=sender)
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\autogen\agentchat\conversable_agent.py", line 781, in generate_reply
final, reply = reply_func(self, messages=messages, sender=sender, config=reply_func_tuple["config"])
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\autogen\agentchat\groupchat.py", line 164, in run_chat
reply = speaker.generate_reply(sender=self)
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\autogen\agentchat\conversable_agent.py", line 781, in generate_reply
final, reply = reply_func(self, messages=messages, sender=sender, config=reply_func_tuple["config"])
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\memgpt\autogen\memgpt_agent.py", line 267, in _generate_reply_for_user_message
) = self.agent.step(user_message, first_message=False, skip_verify=self.skip_verify)
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\memgpt\agent.py", line 655, in step
raise e
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\memgpt\agent.py", line 575, in step
response = self.get_ai_reply(
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\memgpt\agent.py", line 780, in get_ai_reply
raise e
File "C:\Users\Muguntha Pranav M.conda\envs\Auto_mem\lib\site-packages\memgpt\agent.py", line 763, in get_ai_reply
context_window=self.context_window,
AttributeError: 'Agent' object has no attribute 'context_window'

How to solve this error?

@salmercker
Copy link

ok im on windows and no matter how i try to install this using pymemgpt pyautogen or memgpt autogen it doesnt get rid if the lines saying they aint installed

@eldados
Copy link

eldados commented Dec 2, 2023

ok im on windows and no matter how i try to install this using pymemgpt pyautogen or memgpt autogen it doesnt get rid if the lines saying they aint installed

Same here.
On a Mac, Conda environment, installed everything but still get a squiggly line under:
import memgpt.personas.personas as personas
import memgpt.humans.humans as humans

@owgit
Copy link

owgit commented Dec 3, 2023

ok im on windows and no matter how i try to install this using pymemgpt pyautogen or memgpt autogen it doesnt get rid if the lines saying they aint installed

Same here. On a Mac, Conda environment, installed everything but still get a squiggly line under: import memgpt.personas.personas as personas import memgpt.humans.humans as humans

I got same error

@rlam3
Copy link

rlam3 commented Dec 4, 2023

@mberman84
hey mberman84 do you have a new gist that you could share or a repo with a fix?

@marfal
Copy link

marfal commented Dec 6, 2023

Also problems with:
import memgpt.personas.personas as personas
import memgpt.humans.humans as humans

@imfurkaann
Copy link

ModuleNotFoundError: No module named 'memgpt.personas.personas'

how do i solve it

@jakeparkernet
Copy link

jakeparkernet commented Dec 10, 2023 •
edited

Also problems with: import memgpt.personas.personas as personas import memgpt.humans.humans as humans

Also had this, fixed it. If you slow down the video to 0.25x speed, you can see he installs MemGPT 0.1.3 and AutoGen 0.1.13 at the 2:24 timestamp.
I'm running in a virtual environment on Windows. My commands:

pip uninstall pyautogen
pip uninstall pymemgpt
pip install pyautogen==0.1.13
pip install pymemgpt==0.1.3

Successfully ran original gist.

These things are changing and evolving so quickly, it might be a good idea to mention or post versions in future tutorials. Grateful for all the videos just the same.

EDIT: When I back to play the video, it was still 0.25x speed, highly recommend.

@twerpyfie
Copy link

Also problems with: import memgpt.personas.personas as personas import memgpt.humans.humans as humans

Also had this, fixed it. If you slow down the video to 0.25x speed, you can see he installs MemGPT 0.1.3 and AutoGen 0.1.13 at the 2:24 timestamp. I'm running in a virtual environment on Windows. My commands:

pip uninstall pyautogen pip uninstall pymemgpt pip install pyautogen==0.1.13 pip install pymemgpt==0.1.3

Successfully ran original gist.

These things are changing and evolving so quickly, it might be a good idea to mention or post versions in future tutorials. Grateful for all the videos just the same.

EDIT: When I back to play the video, it was still 0.25x speed, highly recommend.

Omg this works, thank you !!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment