Is it possible to make the chatbot remember the previous chat?

[SuhaoCoding]

Most advanced chatbots can remember what the user said and what it responded to, so it will make the new response more related. Is it possible for g4f?

[DarkOperation]

Yes, it is possible. For example, [{"role":"user","content":"hi"},{"role":"assistant","content":"Hello there! How can I assist you today?"},{"role":"user","content":"who are u?"}] is mean, what user sent to model message with conent hi, model sent to user message with content Hello there! How can I assist you today? and user sent to model message with content who are u?.

Here is example:

from g4f.client import Client

client = Client()
messages = []
while True:
    messages.append({"role": "user", "content": input("Enter message to ChatGPT: ")})
    response = client.chat.completions.create(
        model="gpt-3.5-turbo",
        messages=messages,
    )
    messages.append({"role": "assistant", "content": response.choices[0].message.content})

    print(response.choices[0].message.content)

[SuhaoCoding]

Thanks much. But does that mean that the AI will take longer to generate later responses?

[mrtraderbot]

Yes, it is possible. For example, [{"role":"user","content":"hi"},{"role":"assistant","content":"Hello there! How can I assist you today?"},{"role":"user","content":"who are u?"}] is mean, what user sent to model message with conent hi, model sent to user message with content Hello there! How can I assist you today? and user sent to model message with content who are u?.

What about long term memory? Could it easily be setup to store data on each conversation over time?

[microcoder]

Yes! The code with support length of dialogs in bytes and auto reduce:

import g4f
from g4f.client import Client
from g4f.Provider import RetryProvider, Aichatos, FreeGpt
from enum import StrEnum
from IPython.display import display, Markdown, HTML, Pretty


class AssistantRole(StrEnum):
    default = 'You will be a helpful assistant. Answer me in Russian and in Markdown format'
    poet = 'You will be a poet who writes poems and is well versed in rhymes in Russian and responds in Markdown format'
    mathan = 'You will be a helpful assistant that help me with my math homework! Answer me in Russian and in Markdown format'


class GPTChat():


    def __init__(self, model:str='gpt-3.5-turbo-16k-0613', assistant_role=AssistantRole.default, provider=None,
                 context_max_len=None, api_key:str=None, proxies:str=None):
        self.client = Client(
            provider=provider or RetryProvider([Aichatos, FreeGpt], shuffle=False),
            api_key=api_key, proxies=proxies
        )
        self.context = [{'role': 'system', 'content': str(assistant_role)}]
        self.context_max_len = context_max_len or 1024*10
        self.model = model

    @property
    def context_len(self):
        return sum([len(i['content']) for i in self.context[1:]])

    def reduce_context(self):
        '''Reduce owerall length of context to setted context_max_len'''

        while self.context_len > self.context_max_len:
            del self.context[1:3]

    def reset_context(self):
        self.context = self.context[:1]

    def request(self, request:str) -> tuple[str]:

        self.context.append({'role': 'user', 'content': request})

        try:
            response = self.client.chat.completions.create(
                messages=self.context,
                model=self.model,
                stream=False
            )
            True
        except Exception as e:
            self.context.pop()
            raise Exception(e)

        self.context.append({'role': 'assistant', 'content': response.choices[0].message.content})
        self.reduce_context()
        return (self.context[-1]['content'], response.provider, response.model)

Example call:

chat = GPTChat(assistant_role=AssistantRole.mathan, context_max_len=1024*5)

response = chat.request('''
Who are you?
''')

# print(chat.context_max_len, chat.context_len, chat.context)

display(
    Markdown(response[0])
)