app.py

import gradio as gr
from PyPDF2 import PdfReader, PdfFileReader
from langchain.llms import OpenAI
from langchain.schema import Document
from langchain.chains import AnalyzeDocumentChain
from langchain.chains.question_answering import load_qa_chain
from langchain.chat_models import ChatOpenAI
from langchain.callbacks import get_openai_callback
import os


import dotenv
dotenv.load_dotenv('.env')
APP_USER = os.getenv('APP_USER')
APP_PASSWORD = os.environ.get('APP_PASSWORD')

query_template_default= "As you're conducting research in the field of AI, consider the following research question '{question}'. If the answer isn't explicitly found in your current literature review, aim to extrapolate it indirectly. Should the required information be unattainable, please note it as 'Na'. Please remember to cite the sources that influenced your interpretations."

def check_auth(username, password):
    return username == APP_USER and password == APP_PASSWORD

def read_pdf(file):
    try:
        # Use PyPDF2's PdfReader to extract text
        pdf_file_obj = open(file.name, 'rb')
        pdf_reader = PdfReader(pdf_file_obj)
        content = ''
        for page in pdf_reader.pages:
            page_text = page.extract_text()
            if page_text:
                content += page_text
        if not content:
            return "No text found in the PDF."
        return(file.name,content)
    except Exception as e:
        return f"Error reading PDF: {e}"

def query(extracted_text, gpt_model, question,query_template):
    # Process text based on chosen operation
    if gpt_model == "gpt-3.5-turbo":
        llm = ChatOpenAI(temperature=0.7,model="gpt-3.5-turbo")
    elif gpt_model == "gpt-3.5-turbo-16k":
        llm = ChatOpenAI(temperature=0.7,model="gpt-3.5-turbo-16k")
    elif gpt_model == "gpt-4":
        llm = ChatOpenAI(temperature=0.7,model="gpt-4")
    else:
        llm= ChatOpenAI(temperature=0.7,model="gpt-3.5-turbo")
    qa_chain = load_qa_chain(llm, chain_type="map_reduce")
    qa_document_chain = AnalyzeDocumentChain(combine_docs_chain=qa_chain)
    query = query_template.format(question=question)

    #doc = [Document(page_content=extracted_text)]
    with get_openai_callback() as cb:
        answer = qa_document_chain.run(input_document=extracted_text, question=query)
    # Return processed text and feedback
        return [f"Question:\n{question}\nAnser:\n{answer}",f"Total Tokens: {cb.total_tokens}\nPrompt Tokens: {cb.prompt_tokens}\nCompletion Tokens: {cb.completion_tokens}\nTotal Cost (USD): ${cb.total_cost}"]



with gr.Blocks() as ui:
    file_output = gr.File()
    upload_button = gr.UploadButton( label="Upload PDF")
    extracted_text = gr.Text(label="Extracted Text")
    upload_button.upload(read_pdf, upload_button, [file_output,extracted_text])
    query_template = gr.Textbox(label="query_template",value=query_template_default)
    question = gr.Textbox(label="question")
    gpt_model = gr.Radio(choices=["gpt-3.5-turbo", "gpt-3.5-turbo-16k", "gpt-4"], label="Operation")
    result = gr.Text(label="result")
    cost = gr.Text(label="cost")
    query_btn = gr.Button("query")
    query_btn.click(query, inputs=[extracted_text,gpt_model,question,query_template], outputs=[result,cost])

ui.launch(auth=check_auth)