version/semapdf1.4.py

"""
    creator: Lewis Kamau Kimaru
    Function: chat with pdf documents in different languages
    best version yet
"""
from langchain.text_splitter import CharacterTextSplitter
from langchain.embeddings import HuggingFaceBgeEmbeddings
from langchain.vectorstores import FAISS
from langchain.chat_models import ChatOpenAI
from langchain.memory import ConversationBufferMemory
from langchain.chains import ConversationalRetrievalChain
from langchain.llms import HuggingFaceHub

from typing import Union

from dotenv import load_dotenv
from PyPDF2 import PdfReader
import streamlit as st
import requests
import json
import os

# set this key as an environment variable
os.environ["HUGGINGFACEHUB_API_TOKEN"] = st.secrets['huggingface_token']

# Page configuration
st.set_page_config(page_title="SemaNaPDF", page_icon="📚",)

# Sema Translator
Public_Url = 'https://lewiskimaru-helloworld.hf.space' #endpoint

def translate(userinput, target_lang, source_lang=None):
    if source_lang:
       url = f"{Public_Url}/translate_enter/"
       data = {
           "userinput": userinput,
           "source_lang": source_lang,
           "target_lang": target_lang,
        }
       response = requests.post(url, json=data)
       result = response.json()
       print(type(result))
       source_lange = source_lang
       translation = result['translated_text']
       
    else:
      url = f"{Public_Url}/translate_detect/"
      data = {
        "userinput": userinput,
        "target_lang": target_lang,
      }

      response = requests.post(url, json=data)
      result = response.json()
      source_lange = result['source_language']
      translation = result['translated_text']
    return source_lange, translation

def get_pdf_text(pdf : Union[str, bytes, bytearray]) -> str:
    reader = PdfReader(pdf)
    pdf_text = ''
    for page in (reader.pages):
        text = page.extract_text()
        if text:
            pdf_text += text
    return text


def get_text_chunks(text:str) ->list:
    text_splitter = CharacterTextSplitter(
        separator="\n", chunk_size=1500, chunk_overlap=300, length_function=len
    )
    chunks = text_splitter.split_text(text)
    return chunks


def get_vectorstore(text_chunks : list) -> FAISS:
    model = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
    encode_kwargs = {
        "normalize_embeddings": True
    }  # set True to compute cosine similarity
    embeddings = HuggingFaceBgeEmbeddings(
        model_name=model, encode_kwargs=encode_kwargs, model_kwargs={"device": "cpu"}
    )
    vectorstore = FAISS.from_texts(texts=text_chunks, embedding=embeddings)
    return vectorstore


def get_conversation_chain(vectorstore:FAISS) -> ConversationalRetrievalChain:
    llm = HuggingFaceHub(
        repo_id="mistralai/Mixtral-8x7B-Instruct-v0.1",
        #repo_id="TheBloke/Mixtral-8x7B-Instruct-v0.1-GGUF"
        model_kwargs={"temperature": 0.5, "max_length": 1048},
    )

    memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
    conversation_chain = ConversationalRetrievalChain.from_llm(
        llm=llm, retriever=vectorstore.as_retriever(), memory=memory
    )
    return conversation_chain


st.markdown (""" 
  <style> div.stSpinner > div {
    text-align:center; 
    text-align:center;
    align-items: center;
    justify-content: center;
  } 
  </style>""", unsafe_allow_html=True)



def main():
    st.title("SemaNaPDF📚")
    # upload file
    pdf = st.file_uploader("Upload a PDF Document", type="pdf")
    if pdf is not None:
        with st.spinner(""):
            # get pdf text
            raw_text = get_pdf_text(pdf)

            # get the text chunks
            text_chunks = get_text_chunks(raw_text)

            # create vector store
            vectorstore = get_vectorstore(text_chunks)

            # create conversation chain
            st.session_state.conversation = get_conversation_chain(vectorstore)
            st.info("done")
            
    #user_question = st.text_input("chat with your pdf ...")
    # show user input
    if "messages" not in st.session_state:
        st.session_state.messages = []

    for message in st.session_state.messages:
        with st.chat_message(message["role"]):
            st.markdown(message["content"])

    if user_question := st.chat_input("Ask your document anything ......?"):
        with st.chat_message("user"):
            st.markdown(user_question)
            
        user_langd, Queryd = translate(user_question, 'eng_Latn')
        st.session_state.messages.append({"role": "user", "content": user_question})
        response = st.session_state.conversation({"question": Queryd}) #Queryd
        st.session_state.chat_history = response["chat_history"]
        
        output = translate(response['answer'], user_langd, 'eng_Latn')[1] # translated response
        with st.chat_message("assistant"):
            #st.markdown(response['answer'])
            st.markdown(output)
            st.session_state.messages.append({"role": "assistant", "content": response['answer']})

    # Signature
    st.markdown(
        """
        <div style="position: fixed; bottom: 0; right: 0; padding: 10px;">
            <a href="https://kamaukimaru.vercel.app" target="_blank" style="font-size: 12px; color: #269129; text-decoration: none;">©2023 Lewis Kimaru. All rights reserved.</a>
        </div>
        """,
        unsafe_allow_html=True
    )


if __name__ == '__main__':
    main()