app.py

from langchain_core.prompts import PromptTemplate
import os 
from langchain_community.embeddings import HuggingFaceBgeEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_community.llms.ctransformers import CTransformers
from langchain.chains.retrieval_qa.base import RetrievalQA
import streamlit as st
import fitz  # PyMuPDF
from PIL import Image
import io

DB_FAISS_PATH = 'vectorstores/'
#pdf_path = 'data/Harrisons_Internal_Medicine_2022,_21th_Edition_Vol_1_&_Vol_2_.pdf'


custom_prompt_template = '''use the following pieces of information to answer the user's questions.
If you don't know the answer, please just say that don't know the answer, don't try to make up an answer.
Context : {context}
Question : {question}
only return the helpful answer below and nothing else.
'''


def set_custom_prompt():
    """
    Prompt template for QA retrieval for vector stores
    """
    prompt = PromptTemplate(template=custom_prompt_template,
                            input_variables=['context', 'question'])
    return prompt

def load_llm():
    llm = CTransformers(
        #model='epfl-llm/meditron-7b',
        model = 'TheBloke/Llama-2-7B-Chat-GGML',
        model_type='llama',
        max_new_token=512,
        temperature=0.5
    )
    return llm

# def load_embeddings():
#     embeddings = HuggingFaceBgeEmbeddings(model_name='NeuML/pubmedbert-base-embeddings',
#                                           model_kwargs={'device': 'cpu'})
#     return embeddings

# def load_faiss_index(embeddings):
#     db = FAISS.load_local(DB_FAISS_PATH, embeddings, allow_dangerous_deserialization=True)
#     return db

def retrieval_qa_chain(llm, prompt, db):
    qa_chain = RetrievalQA.from_chain_type(
        llm=llm,
        chain_type='stuff',
        retriever=db.as_retriever(search_kwargs={'k': 2}),
        return_source_documents=True,
        chain_type_kwargs={'prompt': prompt}
    )
    return qa_chain

def qa_bot():
    embeddings = HuggingFaceBgeEmbeddings(model_name = 'sentence-transformers/all-MiniLM-L6-v2',
                                          model_kwargs = {'device':'cpu'})
    
    
    db = FAISS.load_local(DB_FAISS_PATH, embeddings, allow_dangerous_deserialization=True)
    llm = load_llm()
    qa_prompt = set_custom_prompt()
    qa = retrieval_qa_chain(llm, qa_prompt, db)
    return qa

def final_result(query):
    qa_result = qa_bot()
    response = qa_result({'query': query})
    return response


def get_pdf_page_as_image(pdf_path, page_number):
    document = fitz.open(pdf_path)
    page = document.load_page(page_number)  
    pix = page.get_pixmap()
    img = Image.open(io.BytesIO(pix.tobytes()))
    return img
    
# # Initialize the bot
# bot = qa_bot()

# Streamlit webpage title
st.title('Medical Chatbot')

# User input
user_query = st.text_input("Please enter your question:")

# Button to get answer
if st.button('Get Answer'):
    if user_query:
        # Call the function from your chatbot script
        response = final_result(user_query)
        if response:
            # Displaying the response
            st.write("### Answer")
            st.write(response['result'])

            # Displaying source document details if available
            if 'source_documents' in response:
                st.write("### Source Document Information")
                for doc in response['source_documents']:
                    # Retrieve and format page content by replacing '\n' with new line
                    formatted_content = doc.page_content.replace("\\n", "\n")
                    st.write("#### Document Content")
                    st.text_area(label="Page Content", value=formatted_content, height=300)

                    # Retrieve source and page from metadata
                    source = doc.metadata['source']
                    page = doc.metadata['page']
                    st.write(f"Source: {source}")
                    st.write(f"Page Number: {page+1}")
                    
                    # Display the PDF page as an image
                    # pdf_page_image = get_pdf_page_as_image(pdf_path, page)
                    # st.image(pdf_page_image, caption=f"Page {page+1} from {source}")
                    
        else:
            st.write("Sorry, I couldn't find an answer to your question.")
    else:
        st.write("Please enter a question to get an answer.")