app.py

import gradio as gr
import os 
import json 
import requests

#Streaming endpoint 
API_URL = "https://api.openai.com/v1/chat/completions" #os.getenv("API_URL") + "/generate_stream"

#Testing with my Open AI Key 
OPENAI_API_KEY = os.getenv("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") 

def predict(inputs, top_p, temperature, openai_api_key, chat_counter, chatbot=[], history=[]):

    # 사용자의 입력을 나레이션 스타일의 프롬프트로 변환
    narration_prompt = f"동영상에 사용할 전문적인 나레이션을 작성하라. 반드시 한글로 작성할것. 일체의 지문이나 지시, 배경 설명 등을 노출 하거나 출력하지 말고 순수한 나레이션만 2줄씩 묶어서 최대 8줄 이내로 출력력. 입력: '{inputs}'"

    payload = {
        "model": "gpt-4-1106-preview",
        "messages": [{"role": "system", "content": narration_prompt}],
        "temperature": temperature,
        "top_p": top_p,
        "n": 1,
        "stream": True,
        "presence_penalty": 0,
        "frequency_penalty": 0,
        "max_tokens": 1000  # 나레이션의 길이를 제한
    }


    headers = {
    "Content-Type": "application/json",
    "Authorization": f"Bearer {openai_api_key}"
    }

    print(f"chat_counter - {chat_counter}")
    if chat_counter != 0 :
        messages=[]
        for data in chatbot:
          temp1 = {}
          temp1["role"] = "user" 
          temp1["content"] = data[0] 
          temp2 = {}
          temp2["role"] = "assistant" 
          temp2["content"] = data[1]
          messages.append(temp1)
          messages.append(temp2)
        temp3 = {}
        temp3["role"] = "user" 
        temp3["content"] = inputs
        messages.append(temp3)
        #messages
        payload = {
        "model": "gpt-4-1106-preview",
        "messages": messages, #[{"role": "user", "content": f"{inputs}"}],
        "temperature" : temperature, #1.0,
        "top_p": top_p, #1.0,
        "n" : 1,
        "stream": True,
        "presence_penalty":0,
        "frequency_penalty":0,
        }

    chat_counter+=1

    history.append(inputs)
    print(f"payload is - {payload}")
    # make a POST request to the API endpoint using the requests.post method, passing in stream=True
    response = requests.post(API_URL, headers=headers, json=payload, stream=True)
    #response = requests.post(API_URL, headers=headers, json=payload, stream=True)
    token_counter = 0 
    partial_words = "" 

    counter=0
    for chunk in response.iter_lines():
        #Skipping first chunk
        if counter == 0:
          counter+=1
          continue
        # 나레이션의 길이를 8줄로 제한
        if len(history) >= 8:
            break            
        #counter+=1
        # check whether each line is non-empty
        if chunk.decode() :
          chunk = chunk.decode()
          # decode each line as response data is in bytes
          if len(chunk) > 12 and "content" in json.loads(chunk[6:])['choices'][0]['delta']:
              #if len(json.loads(chunk.decode()[6:])['choices'][0]["delta"]) == 0:
              #  break
              partial_words = partial_words + json.loads(chunk[6:])['choices'][0]["delta"]["content"]
              if token_counter == 0:
                history.append(" " + partial_words)
              else:
                history[-1] = partial_words
              chat = [(history[i], history[i + 1]) for i in range(0, len(history) - 1, 2) ]  # convert to tuples of list
              token_counter+=1
              yield chat, history, chat_counter  # resembles {chatbot: chat, state: history}  
                

    return chat, history, chat_counter


def reset_textbox():
    return gr.update(value='')

title = """<h1 align="center">혜자 스크립트</h1>"""
description = """영상 생성을 위한 스크립트를 AI가 자동으로 생성합니다. 주제 키워드나 목적 등 필요한 내용만 간단히 입력하세요. :
```
User: <utterance>
Assistant: <utterance>
User: <utterance>
Assistant: <utterance>
...
```
In this app, you can explore the outputs of a gpt-3.5-turbo LLM.
"""


with gr.Blocks(css = """#col_container {width: 1000px; margin-left: auto; margin-right: auto;}
                #chatbot {height: 520px; overflow: auto;}""") as demo:
    gr.HTML(title)
    with gr.Column(elem_id = "col_container"):
        openai_api_key = gr.Textbox(type='password', label="Enter your OpenAI API key here")

        # 출력폼 (chatbot)을 사용자 입력폼 (inputs) 앞에 배치
        chatbot = gr.Chatbot(elem_id='chatbot')  # c
        inputs = gr.Textbox(placeholder="여기에 입력하세요.", label="나레이션 스크립트를 생성하고 싶은 주제어나 문장을 입력하세요.")  # t
        state = gr.State([])  # s
        b1 = gr.Button()
    
        #inputs, top_p, temperature, top_k, repetition_penalty
        with gr.Accordion("Parameters", open=False):
            top_p = gr.Slider( minimum=-0, maximum=1.0, value=1.0, step=0.05, interactive=True, label="Top-p (nucleus sampling)",)
            temperature = gr.Slider( minimum=-0, maximum=5.0, value=1.0, step=0.1, interactive=True, label="Temperature",)
            #top_k = gr.Slider( minimum=1, maximum=50, value=4, step=1, interactive=True, label="Top-k",)
            #repetition_penalty = gr.Slider( minimum=0.1, maximum=3.0, value=1.03, step=0.01, interactive=True, label="Repetition Penalty", )
            chat_counter = gr.Number(value=0, visible=False, precision=0)

    inputs.submit( predict, [inputs, top_p, temperature, openai_api_key, chat_counter, chatbot, state], [chatbot, state, chat_counter],)
    b1.click( predict, [inputs, top_p, temperature, openai_api_key, chat_counter, chatbot, state], [chatbot, state, chat_counter],)
    b1.click(reset_textbox, [], [inputs])
    inputs.submit(reset_textbox, [], [inputs])
                    
    #gr.Markdown(description)
    demo.queue().launch(debug=True)