sd5682295 / EX-chatGPT

Let ChatGPT truly learn how to go online and call APIs! 'EX-ChatGPT' can rival and even surpass NewBing

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

EX-chatGPT introduction

Background

"ChatGPT as Inherent Toolformer" means that ChatGPT has the ability to become a tool for various tasks without requiring additional adjustments.

However, ChatGPT has some limitations such as being unable to connect to the internet and difficulty solving math problems.

ToolFormer enables language models to use specific tools for different tasks. Can ChatGPT be equipped with ToolFormer's abilities?

The challenge is how to adapt ToolFormer's API generation process to ChatGPT.

Recent experiments demonstrate that given a specific prompt, ChatGPT has a natural ability to create APIs for text.

Therefore, it can be concluded that ChatGPT has inherent ToolFormer capabilities!

Toolformer Paper the subproject WebChatGPT enchanced is based on WebChatGPT chrome extension

Demo

ExChatGPT-bilibili API call Demos: API QA Demos: math zhihu zhihu zhihu zhihu

Usage

Ex-chatGPT

  • pip install pip install -r requirements.txt
  • fill your API keys in api_class.py and main.py
    • Googgle api key and client id
    • wolframAlpha api key
    • openAI api key
      • use to summarize when there are too many querys
    • chatGPT access_token
      • (yesterday, OpenAI has published the GPT-3.5 Turbo API , you can replace it with the newsest one)
  • run the main.py and type your query

WebChatGPTEnhance

  • fill you Googgle api key and client id in chatGPTChromeEhance/src/util/apiManager.ts/getDefaultAPI
  • run npm install
  • run npm run build-prod
  • get the extension in chatGPTChromeEhance/build
  • add your prompts and APIs in option page.
    • APIs and prompts examples are in /WebChatGPTAPI
    • wolframAlpha needs to run local sever - WebChatGPTAPI/WolframLocalServer.py

About

Let ChatGPT truly learn how to go online and call APIs! 'EX-ChatGPT' can rival and even surpass NewBing

License:MIT License


Languages

Language:TypeScript 68.2%Language:Python 26.2%Language:JavaScript 4.8%Language:HTML 0.8%Language:CSS 0.1%