app.py

import streamlit as st
import tempfile
import logging
from typing import List
import torch
from langchain_community.document_loaders import PyPDFLoader
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_community.llms import HuggingFacePipeline
from langchain.chains.summarize import load_summarize_chain
from langchain.schema import Document
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.prompts import PromptTemplate
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM

# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Constants
EMBEDDING_MODEL = 'sentence-transformers/all-MiniLM-L6-v2'
DEFAULT_MODEL = "distilgpt2"  # A smaller model that's more likely to work in Spaces

# Check for GPU
device = "cuda" if torch.cuda.is_available() else "cpu"
st.sidebar.write(f"Using device: {device}")

@st.cache_resource
def load_embeddings():
    """Load and cache the embedding model."""
    try:
        return HuggingFaceEmbeddings(model_name=EMBEDDING_MODEL)
    except Exception as e:
        logger.error(f"Failed to load embeddings: {e}")
        st.error("Failed to load the embedding model. Please try again later.")
        return None

@st.cache_resource
def load_llm(model_name):
    """Load and cache the language model."""
    try:
        tokenizer = AutoTokenizer.from_pretrained(model_name)
        model = AutoModelForCausalLM.from_pretrained(model_name)
        pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, device=device, max_length=512)
        return HuggingFacePipeline(pipeline=pipe)
    except Exception as e:
        logger.error(f"Failed to load LLM: {e}")
        st.error(f"Failed to load the model {model_name}. Please try another model or check your internet connection.")
        return None

def process_pdf(file) -> List[Document]:
    """Process the uploaded PDF file."""
    try:
        with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as temp_file:
            temp_file.write(file.getvalue())
            temp_file_path = temp_file.name

        loader = PyPDFLoader(file_path=temp_file_path)
        pages = loader.load()
        text_splitter = RecursiveCharacterTextSplitter(chunk_size=2000, chunk_overlap=100)
        documents = text_splitter.split_documents(pages)
        return documents
    except Exception as e:
        logger.error(f"Error processing PDF: {e}")
        st.error("Failed to process the PDF. Please make sure it's a valid PDF file.")
        return []

def create_vector_store(documents: List[Document], embeddings):
    """Create the vector store."""
    try:
        return FAISS.from_documents(documents, embeddings)
    except Exception as e:
        logger.error(f"Error creating vector store: {e}")
        st.error("Failed to create the vector store. Please try again.")
        return None

def summarize_report(documents: List[Document], llm) -> str:
    """Summarize the report using the loaded model."""
    try:
        prompt_template = """
        Summarize the following text in a clear and concise manner:

        {text}

        Summary:
        """

        prompt = PromptTemplate(template=prompt_template, input_variables=["text"])
        chain = load_summarize_chain(llm, chain_type="stuff", prompt=prompt)
        summary = chain.run(documents)
        return summary

    except Exception as e:
        logger.error(f"Error summarizing report: {e}")
        st.error("Failed to summarize the report. Please try again.")
        return ""

def main():
    st.title("Report Summarizer")

    model_option = st.sidebar.text_input("Enter model name", value=DEFAULT_MODEL)

    uploaded_file = st.sidebar.file_uploader("Upload your Report", type="pdf")

    llm = load_llm(model_option)
    if not llm:
        st.error(f"Failed to load the model {model_option}. Please try another model.")
        return

    embeddings = load_embeddings()
    if not embeddings:
        st.error("Failed to load embeddings. Please try again later.")
        return

    if uploaded_file:
        with st.spinner("Processing PDF..."):
            documents = process_pdf(uploaded_file)

        if documents:
            with st.spinner("Creating vector store..."):
                db = create_vector_store(documents, embeddings)

            if db and st.button("Summarize"):
                with st.spinner(f"Generating summary using {model_option}..."):
                    summary = summarize_report(documents, llm)

                    if summary:
                        st.subheader("Summary:")
                        st.write(summary)
                    else:
                        st.warning("Failed to generate summary. Please try again.")

if __name__ == "__main__":
    main()