app.py

import gradio as gr
from openai import OpenAI
import tiktoken
from os import getenv as os_getenv
from json import loads as json_loads
from pathlib import Path
import fitz

api_key = os_getenv("OPENAI_APIKEY")
client = OpenAI(api_key=api_key)

def get_prompt(books, question = None):
    prompt = (
        f"Read the following books.\n" +
        f"Each book may have some pages at the beggining with data about the book, an index, or table of content, etc. " +
        f"Pages may have a header and/or a footer. Consider all this maybe present." +
        f"Please answer all below in the suggested format, in the language of the book:\n"+
        f"**Title**: ...\n"
        f"**Author**: ...\n"
        f"**Chapter Names**: ...\n"
        f"**Characters**: \n"
        f"**Detailed Summary**: \n"
    )
    prompt += f"{books}\n" 

    return prompt

def chat(message, history, files):
    history_openai_format = []

    if len(history) == 0:
        raise gr.Error("Primero hay que subir un libro")

    if len(history) == 1:
        if message:
            raise gr.Error("First message must be empty")
        message = history[0][0]
    else:
        for human, assistant in history:
            if human:
                history_openai_format.append({"role": "user", "content": human })
            if assistant:
                history_openai_format.append({"role": "assistant", "content":assistant})
    
    history_openai_format.append({"role": "user", "content": message})

    response = client.chat.completions.create(
        model='gpt-4-turbo',
        messages= history_openai_format,
        temperature=1.0,
        stream=True)

    partial_message = ""
    for chunk in response:
        if chunk.choices[0].delta.content is not None:
              partial_message = partial_message + chunk.choices[0].delta.content
              yield partial_message

def get_text(filename):
    answer = ""
    suffix = Path(filename).suffix
    if suffix in [".pdf"]:
        for i,page in enumerate(fitz.open(filename)):
            answer += f"\n### Page #{i+1}\n{page.get_text()}\n"
    elif suffix in [".txt"]:
        answer = open(filename).read()
    return answer

def files_ready(filenames):
    encoder = encoding = tiktoken.encoding_for_model('gpt-4-turbo')
    answer = ''
    for i, name in enumerate(filenames):
        answer += f"\n## Document #{i+1}\nName: {Path(name).name}\n"
        answer += get_text(name)

    return len(encoder.encode(answer)), [[get_prompt(answer), None]]

def files_changed(filenames):
    if not filenames:
        return 0
    
with gr.Blocks(title="Book summarization and more") as demo:
    with gr.Row():
        files = gr.Files(file_types=["txt","doc","docx","pdf"] )
        tokens = gr.Text("0", label="Tokens")

    chat = gr.ChatInterface(
        fn=chat,
        title="Summarization and more",
        additional_inputs=[files],
        multimodal=False)
    
    other = gr.Button(interactive=False)
    files.upload(files_ready, [files], [tokens, chat.chatbot_state])
    files.change(files_changed, files, tokens)


auth=os_getenv("APP_USERS", "null")
auth=json_loads(auth)

demo.launch(auth=auth)