app.py

import streamlit as st
import random, string
import os
import os.path
from os import listdir
from os.path import isfile, join
import requests
import PyPDF2

from langchain.text_splitter import RecursiveCharacterTextSplitter
from pinecone import Pinecone, ServerlessSpec
from groq import Groq
from sentence_transformers import SentenceTransformer


# Access the variables
GROQ_API_KEY = st.secrets["GROQ_API_KEY"]
PINECONE_API_KEY = st.secrets["PINECONE_API_KEY"]
COHERE_API_KEY = st.secrets["COHERE_API_KEY"]

# Initialize Groq client
client = Groq(api_key = GROQ_API_KEY)

# Initialize Pinecone
pc = Pinecone(api_key = PINECONE_API_KEY)

# Create or connect to an existing index
index = pc.Index("sample1")

if 'upload_state' not in st.session_state:
    st.session_state.upload_state = ''

if 'chat_list' not in st.session_state:
    st.session_state.chat_list = []

em_model = SentenceTransformer("all-MiniLM-L6-v2")

def get_query_embdedding(embed):
    query_embedding = em_model.encode([embed]).tolist()   
    return  query_embedding

st.title('Create Summary From Pdf File')

uploaded_files = st.file_uploader("Choose a PDF file", accept_multiple_files=True, type=['pdf'])
# if uploaded_file is not None:
#     bytes_data = uploaded_file.getvalue()
#     data = uploaded_file.getvalue().decode('utf-8', 'ignore').splitlines() 
#     st.session_state["preview"] = ''
#     for i in range(0, min(5, len(data))):
#         st.session_state["preview"] += data[i]
# preview = st.text_area("PDF Preview", "", height=150, key="preview")
# upload_state = st.text_area("Upload State", "", key="upload_state")

pdf_text = ''
ns = "ns_"+''.join(random.choice(string.ascii_uppercase + string.ascii_lowercase + string.digits) for _ in range(7))

def upload(ns):
    for uploaded_file in uploaded_files:
        if uploaded_file is None:
            st.session_state.upload_state = "Upload a file first!"
        else:
            pdf = PyPDF2.PdfReader(uploaded_file)
            pdf_text = ""
            for page in pdf.pages:
                pdf_text += page.extract_text()
            st.session_state.upload_state = "Saved successfully!"

            if pdf_text :
                # Split the text into chunks
                text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
                texts = text_splitter.split_text(pdf_text)

                # # Embedding the chunks
                r1 =  em_model.encode(texts)

                # # Upsert the embeddings into the index
                for i in range(len(texts)):
                    index.upsert([((str(i), r1[i], {"text": texts[i]}))], ns)

    st.write(st.session_state.upload_state)

st.button("Upload file to Process", on_click=upload(ns), key="process_but")
# if st.button("Upload file to Process"):
#     upload(ns)

# Read the PDF file
# pdf = PyPDF2.PdfReader("pdf_files/gandhi.pdf")
# pdf_text = ""
# for page in pdf.pages:
#     pdf_text += page.extract_text()

# Define the query
query = st.chat_input("Enter Your Summarize Query?")
# query = "Who is Bhagat singh?"

docs = ''

if query:
    # Get the query embedding
    question_embedding = get_query_embdedding(query)

    # Query the Pinecone index
    query_result = index.query(namespace = ns, vector=question_embedding, top_k=5, include_metadata=True)

    # Extract metadata from query result
    docs = {x["metadata"]['text']: i for i, x in enumerate(query_result["matches"])}
    # print (docs)     

    # Create a template for the summary
    Template = f"Based on the following context: {docs} generate a precise summary related to the question: {query}"
    # print(Template)   

    # Generate the summary
    chat_completion = client.chat.completions.create(
        messages=[
            {
                "role": "user",
                "content": Template,
            }
        ],
        model="llama3-70b-8192",
    )


    # Print the summary
    response = chat_completion.choices[0].message.content
    # print(response)

    result = {"ques":query, "ans":response}
    st.session_state.chat_list.append(result)

    for c_list in st.session_state.chat_list:
        with st.chat_message("user"):
            st.write(c_list["ques"])
        with st.chat_message("machine"):
            st.write(c_list["ans"])
# if docs:
#     with st.chat_message("chatbot"):
#         st.write(docs)