GPTSimple

GPTSimple is an easier version of the OpenAI python library with more capabilities

Intallation

To get GPTSimple added to your project, download the GPTSimple.py and add it to the directory your project is in.

Usage

import GPTSimple as ai
chat = ai.conversation(base_url="deepinfra") #initialize the conversation
while True: #start the chat loop.
      question = input("> ") #get the user's question
      answer = chat.ask(question) #generate the answer
      print("\n")
      print(answer.text)
      print("\n")

GPTSimple also supports streaming, which lets you get the output before it is done generating

import GPTSimple as ai
chat = ai.conversation(base_url="deepinfra") #initialize the conversation
while True: #start the chat loop.
      question = input("> ") #get the user's question
      print("\n")
      answer = chat.ask(question, stream=True) #generate the answer
      for token in answer: #loop through the tokens in the response
           ai.print_token(token) #print the token
      print("\n")

ai.conversation has many inputs, this is an example of them

#base_url
ai.conversation(base_url="openai")
ai.conversation(base_url="deepinfra")
ai.conversation(base_url="https://api.openai.com/v1/chat/completions")

#api_key (optinal for deepinfra, required for openai)
ai.conversation(api_key="<OpenAI api key>")

#model
ai.conversation(model="gpt-4")
ai.conversation(base_url="deepinfra", model="deepinfra/airoboros-70b")

#system_prompt
ai.conversation(system_prompt="you will taunt the user and make them sad")

#history (overrides system_prompt)
ai.conversation(history=[{"role":"user", "content":"hi"},{"role":"assistant", "content":"i hate you"},{"role":"user", "content":"RUDE!"}])

after you define the conversation, there are two methods to it

#ask
chat.ask("what is your name?") #ask a question and add it to history and get the output
chat.ask("what is your name?", stream=True) #get output as it's generating
chat.ask("what is your name?", invisible=True) #get output without adding it to the history
chat.ask("what is your name?", stream=True, invisible=True)

#generate
chat.generate() #generate the output just from the history, without a question
chat.generate(stream=True) #get output as it's generating
chat.generate(invisible=True) #get output without adding it to the history
chat.generate(stream=True, invisible=True)

after a generation that doesn't use streaming, you will get a response object

response.text #output text from the AI
response.json #exact output from the API
response.message #output text in the {'role':'assistant', 'content':'...'} format
response.output_tokens #amount of tokens in the output
#there are more but i can't find a use for them

If you use streaming, you will get a streamingResponse generator which, when iterated, will return a token object

for token in response: #will be the number of tokens in the response and a extra token to show that it is over
      token.text #content of the token (will be None if the generation is over)
      token.model #will return the model you are using
      token.response #pure response from the API
      token.message #content of the token in the {'role':'assistant', 'content':'...'} format (will be {'role':'assistant', 'content':'None'} if the generation is over)
      ai.print_token(token) #will print the token in a way where you can see it write as it generates

There are a few varibles that might be useful to edit or see

chat.model #change the model (great for compairing)
chat.base_url #only really used along side changing chat.model
chat.api_key
chat.history #set or view the chat history in OpenAI format