├── requirements.txt
├── README.md
└── app.py


/requirements.txt:
--------------------------------------------------------------------------------
1 | streamlit
2 | google-generativeai
3 | python-dotenv
4 | langchain
5 | PyPDF2
6 | chromadb
7 | faiss-cpu
8 | langchain_google_genai


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | # GoogleGemini-PDF
 2 | chat with multiple pdf docs using Google's Gemini generative AI
 3 | 
 4 | In this tutorial, we will explore how to chat with multiple PDF files using Gemini-Pro, a powerful tool that can extract and analyze data from any document. Gemini-Pro is a free software that allows you to interact with your PDF files using natural language queries. You can ask questions, filter results, compare data, and more. Gemini-Pro is easy to use and can handle complex and large PDF files with speed and accuracy.
 5 | 
 6 | ## Required packages:
 7 | https://github.com/Nagh-DE/GoogleGemini-PDF/blob/main/requirements.txt
 8 | 
 9 | ## Main python file 
10 |   (that is used to create front-end, embed the files, and retrieve the responses):
11 | https://github.com/Nagh-DE/GoogleGemini-PDF/blob/main/app.py
12 | 


--------------------------------------------------------------------------------
/app.py:
--------------------------------------------------------------------------------
  1 | import streamlit as st
  2 | from PyPDF2 import PdfReader #library to read pdf files
  3 | from langchain.text_splitter import RecursiveCharacterTextSplitter#library to split pdf files
  4 | import os
  5 | 
  6 | from langchain_google_genai import GoogleGenerativeAIEmbeddings #to embed the text
  7 | import google.generativeai as genai
  8 | 
  9 | from langchain.vectorstores import FAISS #for vector embeddings
 10 | from langchain_google_genai import ChatGoogleGenerativeAI #
 11 | from langchain.chains.question_answering import load_qa_chain #to chain the prompts
 12 | from langchain.prompts import PromptTemplate #to create prompt templates
 13 | from dotenv import load_dotenv
 14 | 
 15 | load_dotenv()
 16 | 
 17 | genai.configure(api_key = os.getenv("GOOGLE_API_KEY"))
 18 | 
 19 | def get_pdf_text(pdf_docs):
 20 |     text = ""
 21 |     # iterate over all pdf files uploaded
 22 |     for pdf in pdf_docs:
 23 |         pdf_reader = PdfReader(pdf)
 24 |         # iterate over all pages in a pdf
 25 |         for page in pdf_reader.pages:
 26 |             text += page.extract_text()
 27 |     return text
 28 | 
 29 | def get_text_chunks(text):
 30 |     # create an object of RecursiveCharacterTextSplitter with specific chunk size and overlap size
 31 |     text_splitter = RecursiveCharacterTextSplitter(chunk_size = 10000, chunk_overlap = 1000)
 32 |     # now split the text we have using object created
 33 |     chunks = text_splitter.split_text(text)
 34 | 
 35 |     return chunks
 36 | 
 37 | def get_vector_store(text_chunks):
 38 |     embeddings = GoogleGenerativeAIEmbeddings(model = "models/embedding-001") # google embeddings
 39 |     vector_store = FAISS.from_texts(text_chunks,embeddings) # use the embedding object on the splitted text of pdf docs
 40 |     vector_store.save_local("faiss_index") # save the embeddings in local
 41 | 
 42 | def get_conversation_chain():
 43 | 
 44 |     # define the prompt
 45 |     prompt_template = """
 46 |     Answer the question as detailed as possible from the provided context, make sure to provide all the details, if the answer is not in
 47 |     provided context just say, "answer is not available in the context", don't provide the wrong answer\n\n
 48 |     Context:\n {context}?\n
 49 |     Question: \n{question}\n
 50 | 
 51 |     Answer:
 52 |     """
 53 | 
 54 |     model = ChatGoogleGenerativeAI(model = "gemini-pro", temperatue = 0.3) # create object of gemini-pro
 55 | 
 56 |     prompt = PromptTemplate(template = prompt_template, input_variables= ["context","question"])
 57 | 
 58 |     chain = load_qa_chain(model,chain_type="stuff",prompt = prompt)
 59 | 
 60 |     return chain
 61 | 
 62 | def user_input(user_question):
 63 |     # user_question is the input question
 64 |     embeddings = GoogleGenerativeAIEmbeddings(model = "models/embedding-001")
 65 |     # load the local faiss db
 66 |     new_db = FAISS.load_local("faiss_index", embeddings)
 67 | 
 68 |     # using similarity search, get the answer based on the input
 69 |     docs = new_db.similarity_search(user_question)
 70 | 
 71 |     chain = get_conversation_chain()
 72 | 
 73 |     
 74 |     response = chain(
 75 |         {"input_documents":docs, "question": user_question}
 76 |         , return_only_outputs=True)
 77 | 
 78 |     print(response)
 79 |     st.write("Reply: ", response["output_text"])
 80 | 
 81 | def main():
 82 |     st.set_page_config("Chat PDF")
 83 |     st.header("Chat with PDF using Gemini")
 84 | 
 85 |     user_question = st.text_input("Ask a Question:")
 86 | 
 87 |     if user_question:
 88 |         user_input(user_question)
 89 | 
 90 |     with st.sidebar:
 91 |         st.title("Menu:")
 92 |         pdf_docs = st.file_uploader("Upload your PDF Files and Click on the Submit & Process Button", accept_multiple_files=True)
 93 |         if st.button("Submit & Process"):
 94 |             with st.spinner("Processing..."):
 95 |                 raw_text = get_pdf_text(pdf_docs)
 96 |                 text_chunks = get_text_chunks(raw_text)
 97 |                 get_vector_store(text_chunks)
 98 |                 st.success("Done")
 99 | 
100 | 
101 | if __name__ == "__main__":
102 |     main()
103 | 


--------------------------------------------------------------------------------