core.py

import asyncio
import textwrap

import openai
from halo import Halo

gpt_costs_per_thousand_out = {
    "gpt-3.5-turbo-16k": 0.004,
    "gpt-4-32k": 0.12,
}
gpt_costs_per_thousand_in = {
    "gpt-3.5-turbo-16k": 0.003,
    "gpt-4-32k": 0.06,
}


def estimate_costs(prompt_tokens, model: str):
    costs = (prompt_tokens / 1000) * gpt_costs_per_thousand_in[model]
    return costs


async def chatbot(conversation, model, temperature=0):
    max_retry = 7
    retry = 0
    while True:
        try:
            response = await openai.ChatCompletion.acreate(
                model=model, messages=conversation, temperature=temperature
            )
            text = response["choices"][0]["message"]["content"]

            return text, response["usage"]
        except Exception as oops:
            print(f'\n\nError communicating with OpenAI: "{oops}"')
            if "maximum context length" in str(oops):
                a = conversation.pop(0)
                print("\n\n DEBUG: Trimming oldest message")
                continue
            retry += 1
            if retry >= max_retry:
                print(f"\n\nExiting due to excessive errors in API: {oops}")
                exit(1)
            print(f"\n\nRetrying in {2 ** (retry - 1) * 5} seconds...")
            await asyncio.sleep(2 ** (retry - 1) * 5)


def chat_print(text):
    formatted_lines = [
        textwrap.fill(line, width=120, initial_indent="    ", subsequent_indent="    ")
        for line in text.split("\n")
    ]
    formatted_text = "\n".join(formatted_lines)
    print("\n\n\nCHATBOT:\n\n%s" % formatted_text)