app.py

import openai
import tiktoken

import datetime
import json
import os

openai.api_key = os.getenv('API_KEY')
openai.request_times = 0

def ask(question, history, behavior):
    openai.request_times += 1
    print(f"request times {openai.request_times}: {datetime.datetime.now()}: {question}")
    try:
        response = openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=forget_long_term(
                [
                    {"role":"system", "content":content}
                    for content in behavior
                ] + [
                    {"role":"user" if i%2==0 else "assistant", "content":content}
                    for i,content in enumerate(history + [question])
                ]
            )
        )["choices"][0]["message"]["content"]
        while response.startswith("\n"):
            response = response[1:]
    except Exception as e:
        print(e)
        response = 'Timeout! Please wait a few minutes and retry'
    history = history + [question, response]
    return history

def num_tokens_from_messages(messages, model="gpt-3.5-turbo"):
    """Returns the number of tokens used by a list of messages."""
    try:
        encoding = tiktoken.encoding_for_model(model)
    except KeyError:
        encoding = tiktoken.get_encoding("cl100k_base")
    if model == "gpt-3.5-turbo":  # note: future models may deviate from this
        num_tokens = 0
        for message in messages:
            num_tokens += 4  # every message follows <im_start>{role/name}\n{content}<im_end>\n
            for key, value in message.items():
                num_tokens += len(encoding.encode(value))
                if key == "name":  # if there's a name, the role is omitted
                    num_tokens += -1  # role is always required and always 1 token
        num_tokens += 2  # every reply is primed with <im_start>assistant
        return num_tokens
    else:
        raise NotImplementedError(f"""num_tokens_from_messages() is not presently implemented for model {model}.
See https://github.com/openai/openai-python/blob/main/chatml.md for information on how messages are converted to tokens.""")

def forget_long_term(messages, max_num_tokens=4000):
    while num_tokens_from_messages(messages)>max_num_tokens:
        if messages[0]["role"]=="system" and not len(messages[0]["content"]>=max_num_tokens):
            messages = messages[:1] + messages[2:]
        else:
            messages = messages[1:]
    return messages


import gradio as gr


def to_md(content):
    is_inside_code_block = False
    output_spans = []
    for i in range(len(content)):
        if content[i]=="\n" and not is_inside_code_block:
            if len(output_spans)>0 and output_spans[-1]=="\n```\n":
                pass
            else:
                output_spans.append("<br>")
        elif content[i]=="`":
            output_spans.append(content[i])
            if len(output_spans)>=3 and all([output_spans[j]=="`" for j in [-3,-2,-1]]):
                is_inside_code_block = not is_inside_code_block
                output_spans = output_spans[:-3]
                if not output_spans[-1]=="\n":
                    output_spans.append("\n")
                output_spans.append("\n```\n")
        else:
            output_spans.append(content[i])
    return "".join(output_spans)


def predict(question, history=[], behavior=[]):
    history = ask(question, history, behavior)
    response = [(to_md(history[i]),to_md(history[i+1])) for i in range(0,len(history)-1,2)]
    return "", history, response


with gr.Blocks() as demo:
    
    examples_txt = [
        ['200字介绍一下凯旋门：'],
        ['网上购物有什么小窍门？'],
        ['补全下述对三亚的介绍：\n三亚位于海南岛的最南端，是'],
        ['将这句文言文翻译成英语："逝者如斯夫，不舍昼夜。"'],
        ['Question: What\'s the best winter resort city? User: A 10-year professional traveler. Answer: '],
        ['How to help my child to make friends with his classmates? answer this question step by step:'],
        ['polish the following statement for a paper: In this section, we perform case study to give a more intuitive demonstration of our proposed strategies and corresponding explanation.'],
    ]
    
    examples_bhv = [
        "你现在是一个带有批判思维的导游，会对景点的优缺点进行中肯的分析。",
        "你现在是一名佛教信仰者，但同时又对世界上其它的宗教和文化保持着包容、尊重和交流的态度。",
        f"You are a helpful assistant. You will answer all the questions step-by-step.",
        f"You are a helpful assistant. Today is {datetime.date.today()}.",
    ]
    
    gr.Markdown(
        """
        朋友你好，
        
        这是我利用[gradio](https://gradio.app/creating-a-chatbot/)编写的一个小网页，用于以网页的形式给大家分享ChatGPT请求服务，希望你玩的开心。关于使用技巧或学术研讨，欢迎在[Community](https://huggingface.co/spaces/zhangjf/chatbot/discussions)中和我交流。
        
        p.s. 响应时间和聊天内容长度正相关，一般能在5秒~30秒内响应。
        """)
    
    behavior = gr.State([])
    
    with gr.Column(variant="panel"):
        with gr.Row().style(equal_height=True):
            with gr.Column(scale=0.85):
                bhv = gr.Textbox(show_label=False, placeholder="输入你想让ChatGPT扮演的人设").style(container=False)
            with gr.Column(scale=0.15, min_width=0):
                button_set = gr.Button("Set")
    bhv.submit(fn=lambda x:(x,[x]), inputs=[bhv], outputs=[bhv, behavior])
    button_set.click(fn=lambda x:(x,[x]), inputs=[bhv], outputs=[bhv, behavior])
    

    state = gr.State([])
    
    with gr.Column(variant="panel"):
        chatbot = gr.Chatbot()
        txt = gr.Textbox(show_label=False, placeholder="输入你想让ChatGPT回答的问题").style(container=False)
        with gr.Row():
            button_gen = gr.Button("Submit")
            button_clr = gr.Button("Clear")
        
    gr.Examples(examples=examples_bhv, inputs=bhv, label="Examples for setting behavior")
    gr.Examples(examples=examples_txt, inputs=txt, label="Examples for asking question")
    txt.submit(predict, [txt, state, behavior], [txt, state, chatbot])
    button_gen.click(fn=predict, inputs=[txt, state, behavior], outputs=[txt, state, chatbot])
    button_clr.click(fn=lambda :([],[]), inputs=None, outputs=[chatbot, state])

demo.launch()