pyChat

This project tries to be a free library to create a python assistant with GPT4 and other providers like Bing Search, Watson Speech to Text, Bing Speech to Text, Google Text to Speech and more.

It's a python library that can be used to import in other projects, without any API key, and with free internet partners.

Models engine implemented

Provider	GPT3.5	GPT4	C. Memory	I. Browser	Stream	Working	In GUI	Notes
OpenChat	✘	✔	✔	✘	✔	✔	✔	Sometimes online services is down for maintenance
Bing	✔	✔	✔	✔	✔	✔	✔	Automatized captcha solver with pySide6 GUI (Desktop), and prompt (last one needs Firefox and Linux)
gpt4free	✔	✘	✔	✘	✔	✔	✔	His engine has changed, it's consuming gptgod.site, so probably I will append his provider and remove this one
Wrtn	✔	✔	✘	✘	✔	✔	✔	South Korean by default (so you have to require in your command the answer in your language if you don't want to read in Korean)
ChatGPTSpanish	✔	✔	✘	✘	✘	✔	✔	You're able to ask one time per petition, but you can clean and ask again
You	✔	✔	✔	✔	✔	✔	✔	GPT-4 promps needs login and are limited to 5 per account, but supports auto register a random account and login
ch4	✔	✔	✔	✘	✔	✔	✘	It's too slow, even streaming, but works. Needs to be logged in (you've auto-registration, but it's preferible use always the same cookie). Sometimes platform forces using Gemini-Pro, but... it's a platform bug.

Speech to text engines

Provider	Working	In GUI
Watson	✔	✘
Bing	✔	✔

Text to speech engine

Provider	Working	In GUI
Google	✔	✘

How to use it (future wiki section)

There is a pytest file that could be used to test the library. Anyway you can use it for your own tests, for instance:

realPetitionPromptNew = "¿qué modelo de lenguaje estás utilizando? ¿chatgpt3 o chatgpt4?"

# You
from core.you import You
you=You()
you.send_message(realPetitionPromptNew, stream=True)

# openchat
from core.openchat import OpenChat
openchat = OpenChat()
openchat.send_message(realPetitionPromptNew, stream=True)

# wrtn.ai
from core.wrtnai import WRTNAI
wrtnai = WRTNAI()
wrtnai.prompt(realPetitionPromptNew)

# gpt4free.io/chat
from core.gpt4free import Gpt4free
gpt4free=Gpt4free()
gpt4free.prompt(realPetitionPromptNew)

# chatgptspanish
from core.chatgptspanish import ChatGPTSpanish
chatgptspanish = ChatGPTSpanish()
chatgptspanish.send_message(realPetitionPromptNew)

# ch4
from core.ch4usto import Ch4usto
c=Ch4usto()
c.send_message(realPetitionPromptNew)

# Bing
from core.bing import Bing
bing = Bing()
# Bing speech to text
bing.speech_to_text()
# BingGPT AI
bing.init_conversation(realPetitionPromptNew)

# Watson speech to text
from core.watson import Watson
watson = Watson()
watson.speech_to_text()

# Google text to speech
from core.translator import Translator
translator = Translator()
translator.play(realPetitionPromptNew)

And enjoy it!

Streams

In this kind of APIs you will concern the streaming ability to get a real time response. In this library with the partners which provide streaming you have to add a queue:

import queue
myQueue = queue.Queue()

and use it:

from core.wrtnai import WRTNAI
wrtnai = WRTNAI()
wrtnai.prompt(cmd=realPetitionPromptNew, queue=myQueue, stream=True)

Queue could be read from outside if you consider use asyncio implementation.

Developer notes

Why?

I always want to know how the things works, and other libraries don't show you how the things works.

I want to offer an updated library with some free models, to use it for free, and all will be open source.

GUI frameworks

Now I'm centered in core functionalities, but I've tried to develop a multiplatform GUI:

Currently I'm working on poetry and pySide6 multiplatform GUI. To build it run:

curl -sSL https://install.python-poetry.org | python -
python -m poetry shell
pip install -r requirements.txt
poetry update
poetry install
poetry run gui

But if you want to make all in two commands, make it easy:

pip install -r requirements.txt
python main.py

And enjoy.

dev Briefcase issues

In the past I used Briefcase with Toga. It was an awesome project but Briefcase/Toga/Gbulb-GTK has some with aiohttp and asyncio calls in GTK (related to his subproject gbulb, that it has not implement very well wss throw ssl, and toga-gtk, forgiven too many time for developers), so I discarted it officially. I tried to fix it, without success, and I'm not the first one ). So, multiplatform GUI was implemented but unsupported with briefcase.

With a working mobile version:

If you want know about how to use it with Toga/Briefcase take a look about this commit in the README.md file.

PIP issues

In ArchLinux pip is configured to be used with --break-system-packages or system, buildozer needs a fix. Edit ~/.config/pip/pip.conf

[global]
break-system-packages = true

And enjoy it!

License

This project is licensed under the terms of the CC BY-NC-ND 4.0 license, and released by @bitstuffing with love.

It's a open development in an investigation phase, and it's not allowed for business use.