app.py

import os
import requests
import streamlit as st
import PyMuPDF

# Get the Hugging Face API Token from environment variables
HF_API_TOKEN = os.getenv("HF_API_KEY")
if not HF_API_TOKEN:
    raise ValueError("Hugging Face API Token is not set in the environment variables.")

# Hugging Face API URL and header for Gemma 27B-it model
GEMMA_27B_API_URL = "https://api-inference.huggingface.co/models/google/gemma-2-27b-it"

HEADERS = {"Authorization": f"Bearer {HF_API_TOKEN}"}

def query_model(api_url, payload):
    response = requests.post(api_url, headers=HEADERS, json=payload)
    return response.json()

def extract_pdf_text(uploaded_file):
    pdf_text = ""
    pdf_doc = PyMuPDF.open(uploaded_file)
    for page_num in range(len(pdf_doc)):
        pdf_text += pdf_doc.getPageText(page_num)
    return pdf_text

def add_message_to_conversation(user_message, bot_message, model_name):
    if "conversation" not in st.session_state:
        st.session_state.conversation = []
    st.session_state.conversation.append((user_message, bot_message, model_name))

# Streamlit app
st.set_page_config(page_title="Gemma 27B-it Chatbot Interface", layout="wide")
st.title("Gemma 27B-it Chatbot Interface")
st.write("Gemma 27B-it Chatbot Interface")

# Initialize session state for conversation and uploaded file
if "conversation" not in st.session_state:
    st.session_state.conversation = []
if "uploaded_file" not in st.session_state:
    st.session_state.uploaded_file = None

# File uploader for PDF
uploaded_file = st.file_uploader("Upload a PDF", type="pdf")

# Handle PDF upload and text extraction
if uploaded_file:
    pdf_text = extract_pdf_text(uploaded_file)
    st.write("### PDF Text Extracted:")
    st.write(pdf_text)

# User input for question
question = st.text_input("Question", placeholder="Enter your question here...")

# Handle user input and Gemma 27B-it model response
if st.button("Send") and question:
    try:
        with st.spinner("Waiting for the model to respond..."):
            # Construct the chat history
            chat_history = " ".join([msg[1] for msg in st.session_state.conversation[-5:]]) + f"User: {question}\n"
            response = query_model(GEMMA_27B_API_URL, {"inputs": chat_history})

            if isinstance(response, list):
                answer = response[0].get("generated_text", "No response")
            elif isinstance(response, dict):
                answer = response.get("generated_text", "No response")
            else:
                answer = "No response"

            # Add PDF text to the chat history
            if st.session_state.uploaded_file:
                chat_history += f"Document Text: {pdf_text}\n"

            add_message_to_conversation(question, answer, "Gemma-2-27B-it")
    except ValueError as e:
        st.error(str(e))

# Custom CSS for chat bubbles
st.markdown(
    """
    <style>
    .chat-bubble {
        padding: 10px 14px;
        border-radius: 14px;
        margin-bottom: 10px;
        display: inline-block;
        max-width: 80%;
        color: black;
    }
    .chat-bubble.user {
        background-color: #dcf8c6;
        align-self: flex-end;
    }
    .chat-bubble.bot {
        background-color: #fff;
        align-self: flex-start;
    }
    .chat-container {
        display: flex;
        flex-direction: column;
        gap: 10px;
        margin-top: 20px;
    }
    </style>
    """,
    unsafe_allow_html=True
)

# Display the conversation
st.write('<div class="chat-container">', unsafe_allow_html=True)
for user_message, bot_message, model_name in st.session_state.conversation:
    st.write(f'<div class="chat-bubble user">You: {user_message}</div>', unsafe_allow_html=True)
    st.write(f'<div class="chat-bubble bot">{model_name}: {bot_message}</div>', unsafe_allow_html=True)
st.write('</div>', unsafe_allow_html=True)