openai_o1_mini_preview_streaming.py

import openai
import os
import sys
import inspect
import ast
import difflib

# Initialize OpenAI client
openai.api_type = 'openai'
openai.api_key = os.getenv("OPENAI_API_KEY")
client = openai.OpenAI()

# Define models
MODELS = {
    'o1-preview': 'o1-preview',
    'o1-mini': 'o1-mini',
    'gpt-4o': 'gpt-4o',
    'gpt-4o-mini': 'gpt-4o-mini'
}

def o1_preview(user_prompt: str = "", system_prompt: str = "", context: str = "", messages: Optional[List[Dict[str, Any]]] = []):
    model = MODELS['o1-preview']
    
    if not messages or len(messages) == 0:
      messages = [
          {"role": "user", "content": f"""<system_prompt> {system_prompt} </system_prompt>
  <context> {context} </context>
  <user_prompt> {user_prompt } </user_prompt>"""
          }
      ]
    response = client.chat.completions.create(model=model, messages=messages, stream=True)
    return response

def o1_mini(user_prompt: str = "", system_prompt: str = "", context: str = "", messages: Optional[List[Dict[str, Any]]] = []):
    
    model = MODELS['o1-mini']
    if not messages or len(messages) == 0:
      messages = [
          {"role": "user", "content": f"""<system_prompt> {system_prompt} </system_prompt>
  <context> {context} </context>
  <user_prompt> {user_prompt } </user_prompt>"""
          }
      ]
    
    response = client.chat.completions.create(model=model, messages=messages, stream=True)
    return response

def gpt_4o(messages, **kwargs):
    model = MODELS['gpt-4o']
    response = client.chat.completions.create(model=model, messages=messages, **kwargs)
    return response

def gpt_4o_mini(messages, **kwargs):
    model = MODELS['gpt-4o-mini']
    response = client.chat.completions.create(model=model, messages=messages, **kwargs)
    return response
 
SYSTEM_PROMPT = "You are the smartest intelligence of its kind, the precursor"

def o1(prompt: str, system_prompt: str = SYSTEM_PROMPT):
    return o1_preview(
        user_prompt=prompt,
        system_prompt='I am an otherworldly AI.'
    )
 
def stream_response_to_cli(response_generator):
    """
    Streams the tokens from a response generator to the CLI.

    Args:
        response_generator: An iterable that yields response chunks.
    """
    assistant_message = ''
    for chunk in response_generator:
        # Extract the content from each chunk
        content = chunk.choices[0].delta.content
        # Print the content without adding a newline to stream the tokens
        if content:
            assistant_message += content
        print(content, end='', flush=True)
    print()  # Print a final newline after completion
    return {"role": "assistant", "content": assistant_message}

text = 'import openai\nimport os\nimport sys\nimport inspect\nimport ast\nimport difflib\n\n# Initialize OpenAI client\nopenai.api_type = \'openai\'\nopenai.api_key = os.getenv("OPENAI_API_KEY")\nclient = openai.OpenAI()\n\n# Define models\nMODELS = {\n    \'o1-preview\': \'o1-preview\',\n    \'o1-mini\': \'o1-mini\',\n    \'gpt-4o\': \'gpt-4o\',\n    \'gpt-4o-mini\': \'gpt-4o-mini\'\n}\n\ndef o1_preview(user_prompt: str = "", system_prompt: str = "", context: str = "", messages: Optional[List[Dict[str, Any]]] = []):\n    model = MODELS[\'o1-preview\']\n\n    if not messages or len(messages) == 0:\n      messages = [\n          {"role": "user", "content": f\'<system_prompt> {system_prompt} </system_prompt>\n<context> {context} </context>\n<user_prompt> {user_prompt } </user_prompt>\'\n          }\n      ]\n    response = client.chat.completions.create(model=model, messages=messages, stream=True)\n    return response\n\ndef o1_mini(user_prompt: str = "", system_prompt: str = "", context: str = "", messages: Optional[List[Dict[str, Any]]] = []):\n\n    model = MODELS[\'o1-mini\']\n    if not messages or len(messages) == 0:\n      messages = [\n          {"role": "user", "content": f\'<system_prompt> {system_prompt} </system_prompt>\n<context> {context} </context>\n<user_prompt> {user_prompt } </user_prompt>\'\n          }\n      ]\n\n    response = client.chat.completions.create(model=model, messages=messages, stream=True)\n    return response\n\ndef gpt_4o(messages, **kwargs):\n    model = MODELS[\'gpt-4o\']\n    response = client.chat.completions.create(model=model, messages=messages, **kwargs)\n    return response\n\ndef gpt_4o_mini(messages, **kwargs):\n    model = MODELS[\'gpt-4o-mini\']\n    response = client.chat.completions.create(model=model, messages=messages, **kwargs)\n    return response\n\nSYSTEM_PROMPT = "You are the smartest intelligence of its kind, the precursor"\n\ndef o1(prompt: str, system_prompt: str = SYSTEM_PROMPT, stream: bool=True):\n    return o1_preview(prompt, system_prompt, stream=stream)\n    )\n\ndef stream_response_to_cli(response_generator):\n    assistant_message = \'\'\n    for chunk in response_generator:\n        # Extract the content from each chunk\n        content = chunk.choices[0].delta.content\n        # Print the content without adding a newline to stream the tokens\n        if content:\n            assistant_message += content\n        print(content, end=\'\', flush=True)\n    print()  # Print a final newline after completion\n    return {"role": "assistant", "content": assistant_message}\n\nresponse_generator = o1_preview(\n    user_prompt="<system_prompt> You will iteratively rewrite your system prompt as we progress through the convo. Before each message you will rewrite the system prompt based on what you\'re learning. You will use a little agent scratchpad to maintain state </system_prompt> <user prompt> Let\'s think slowly. Step by step. What is the best way to make a meta-recursive self-aware transformer architecture? Try to include specific code segments, specific areas to look at in terms of modifying a typical Llama 2/3/3.1 style architecture for these additional capabilities </user_prompt>"\n);\n\nstream_response_to_cli(response_generator)'

response_generator = o1_preview(user_prompt="Rewrite this as a streaming FastAPI app!", system_prompt="The o1-preview and o1-mini models were just released. Today, we will write a converastional CLI chat where we create a convo, it gets a UUID and an empty list of messages. I pass in system_prompt and it really just gets added into the user message, but please write an alterantive chat, assistant and user, let's store messages in a JSON file. Rewrite comprehensively", context=text)

stream_response_to_cli(response_generator)