app.py

import gradio as gr
import os
import time
import threading
from langchain.document_loaders import OnlinePDFLoader
from langchain.text_splitter import CharacterTextSplitter
from langchain.llms import OpenAI
from langchain.embeddings import OpenAIEmbeddings
from langchain.vectorstores import Chroma
from langchain.chains import ConversationalRetrievalChain
from langchain.chat_models import ChatOpenAI
from langchain.document_loaders import WebBaseLoader
from langchain.chains.summarize import load_summarize_chain
from langchain.chains.llm import LLMChain
from langchain.prompts import PromptTemplate
from langchain.chains.combine_documents.stuff import StuffDocumentsChain

os.environ['OPENAI_API_KEY'] = os.getenv("Your_API_Key")

# Global variable for tracking last interaction time
last_interaction_time = 0

def loading_pdf():
    return "Working on the upload. Also, pondering the usefulness of sporks..."

# Inside Chroma mod
def summary(self):
    num_documents = len(self.documents)
    avg_doc_length = sum(len(doc) for doc in self.documents) / num_documents
    return f"Number of documents: {num_documents}, Average document length: {avg_doc_length}"

# Gradio state
summary_state = gr.State(initial_value="pending")

# PDF summary and query using stuffing
def pdf_changes(pdf_doc):
    try:
        # Initialize loader and load documents
        loader = OnlinePDFLoader(pdf_doc.name)
        documents = loader.load()

        # Define the prompt for summarization
        prompt_template = """Write a concise summary of the following:
        "{text}"
        CONCISE SUMMARY:"""
        prompt = PromptTemplate.from_template(prompt_template)

        # Define the LLM chain with the specified prompt
        llm = ChatOpenAI(temperature=0, model_name="gpt-3.5-turbo-16k")
        llm_chain = LLMChain(llm=llm, prompt=prompt)

        # Initialize StuffDocumentsChain
        stuff_chain = StuffDocumentsChain(
            llm_chain=llm_chain, document_variable_name="text"
        )

        # Generate summary using StuffDocumentsChain
        global full_summary
        full_summary = stuff_chain.run(documents)
        # Update the state variable
        return {summary_state: full_summary}

        # Other existing logic for Chroma, embeddings, and retrieval
        embeddings = OpenAIEmbeddings()
        global db
        db = Chroma.from_documents(documents, embeddings)

        retriever = db.as_retriever()
        global qa
        qa = ConversationalRetrievalChain.from_llm(
            llm=OpenAI(temperature=0.2, model_name="gpt-3.5-turbo-16k", max_tokens=-1, n=2),
            retriever=retriever,
            return_source_documents=False
        )
        summary_box.set_value(full_summary)
        return f"Ready. Full Summary loaded."

    except Exception as e:
        return f"Error processing PDF: {str(e)}"



def clear_data():
    global qa, db
    qa = None
    db = None
    return "Data cleared"

def add_text(history, text):
    global last_interaction_time
    last_interaction_time = time.time()
    history = history + [(text, None)]
    return history, ""

def bot(history):
    global full_summary  
    if 'summary' in history[-1][0].lower():  # Check if the last question asks for a summary
        response = full_summary
        return full_summary
    else:
        response = infer(history[-1][0], history)

    sentences = '  \n'.join(response.split('. '))
    formatted_response = f"**Bot:**\n\n{sentences}"
    history[-1][1] = formatted_response
    return history


def infer(question, history):
    try:
        res = []
        for human, ai in history[:-1]:
            pair = (human, ai)
            res.append(pair)
    
        chat_history = res
        query = question
        result = qa({"question": query, "chat_history": chat_history, "system": "This is a world-class summarizing AI, be helpful."})
        return result["answer"]
    except Exception as e:
        return f"Error querying chatbot: {str(e)}"

def auto_clear_data():
      global qa, db, last_interaction_time
      if time.time() - last_interaction_time > 1000:
        qa = None
        db = None
        print("Data cleared successfully.")  # Logging
        
def periodic_clear():
      while True:
        auto_clear_data()
        time.sleep(1000)
        
threading.Thread(target=periodic_clear).start()
        
css = """
#col-container {max-width: 700px; margin-left: auto; margin-right: auto;}
"""
        
title = """
<div style="text-align: center;max-width: 700px;">
    <h1>CauseWriter Chat with PDF • OpenAI</h1>
    <p style="text-align: center;">Upload a .PDF from your computer, click the "Load PDF to LangChain" button, <br />
    when everything is ready, you can start asking questions about the pdf. Limit ~11k words. <br />
    This version is set to erase chat history automatically after page timeout and uses OpenAI.</p>
</div>
"""
        
with gr.Blocks(css=css) as demo:
        with gr.Column(elem_id="col-container"):
            gr.HTML(title)
                
            with gr.Column():
                  pdf_doc = gr.File(label="Load a pdf", file_types=['.pdf'], type="file")
                  with gr.Row():
                      langchain_status = gr.Textbox(label="Status", placeholder="", interactive=False)
                      load_pdf = gr.Button("Convert PDF to Magic AI language")
                      clear_btn = gr.Button("Clear Data")
                    
                    # New Textbox to display summary
                  summary_box = gr.Textbox(
                      label="Document Summary",
                      placeholder="Summary will appear here.",
                      interactive=False,
                      rows=5,
                      elem_id="summary_box"  # Set the elem_id to match the state key
                    )
                
            chatbot = gr.Chatbot([], elem_id="chatbot").style(height=450)
            question = gr.Textbox(label="Question", placeholder="Type your question and hit Enter")
            submit_btn = gr.Button("Send Message")
        
            # Step 2 and 3: Put the State object as an input and output
            load_pdf.click(pdf_changes, inputs=[pdf_doc, summary_state], outputs=[langchain_status, summary_state])
            clear_btn.click(clear_data, outputs=[langchain_status])
            question.submit(add_text, [chatbot, question], [chatbot, question]).then(
                  bot, chatbot, chatbot
            )
            submit_btn.click(add_text, [chatbot, question], [chatbot, question]).then(
                  bot, chatbot, chatbot
            )
        
demo.launch()