app.py

import streamlit as st
import os
import tempfile
from typing import List, Optional
import pickle

# Core libraries
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline
from langchain.llms import HuggingFacePipeline
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.schema import Document
from langchain import PromptTemplate
from langchain.chains import RetrievalQA
from langchain.vectorstores import FAISS

# Document loaders
from langchain.document_loaders import PyPDFLoader

# Configure Streamlit page
st.set_page_config(
    page_title="PDF RAG System",
    page_icon="📚",
    layout="wide",
    initial_sidebar_state="expanded"
)

# Custom CSS for better styling
st.markdown("""
<style>
    .main-header {
        font-size: 2.5rem;
        color: #1f77b4;
        text-align: center;
        margin-bottom: 2rem;
    }
    .sidebar-header {
        font-size: 1.5rem;
        color: #ff7f0e;
        margin-bottom: 1rem;
    }
    .success-message {
        padding: 1rem;
        background-color: #d4edda;
        border: 1px solid #c3e6cb;
        border-radius: 0.5rem;
        color: #155724;
        margin: 1rem 0;
    }
    .error-message {
        padding: 1rem;
        background-color: #f8d7da;
        border: 1px solid #f5c6cb;
        border-radius: 0.5rem;
        color: #721c24;
        margin: 1rem 0;
    }
    .source-box {
        background-color: #f8f9fa;
        border-left: 4px solid #007bff;
        padding: 1rem;
        margin: 0.5rem 0;
        border-radius: 0 0.5rem 0.5rem 0;
    }
</style>
""", unsafe_allow_html=True)

# Initialize session state
if 'qa_chain' not in st.session_state:
    st.session_state.qa_chain = None
if 'vectorstore' not in st.session_state:
    st.session_state.vectorstore = None
if 'documents_processed' not in st.session_state:
    st.session_state.documents_processed = False
if 'chat_history' not in st.session_state:
    st.session_state.chat_history = []

@st.cache_resource
def setup_llm(model_name="google/flan-t5-small"):
    """Setup the language model for text generation"""
    with st.spinner("🤖 Loading language model..."):
        try:
            tokenizer = AutoTokenizer.from_pretrained(model_name)
            model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
            
            pipe = pipeline(
                "text2text-generation",
                model=model,
                tokenizer=tokenizer,
                max_new_tokens=300,
                temperature=0.3,
                do_sample=True,
                device=-1
            )
            
            llm = HuggingFacePipeline(pipeline=pipe)
            return llm
        except Exception as e:
            st.error(f"Error loading model: {e}")
            return None

@st.cache_resource
def setup_embeddings(model_name="all-MiniLM-L6-v2"):
    """Setup the embedding model for vector generation"""
    with st.spinner("🔢 Loading embedding model..."):
        try:
            embeddings = HuggingFaceEmbeddings(model_name=model_name)
            return embeddings
        except Exception as e:
            st.error(f"Error loading embeddings: {e}")
            return None

def process_uploaded_files(uploaded_files, embeddings):
    """Process uploaded PDF files and create FAISS vector store"""
    if not uploaded_files:
        return None, []
    
    documents = []
    
    # Process each uploaded file
    for uploaded_file in uploaded_files:
        try:
            # Create temporary file
            with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp_file:
                tmp_file.write(uploaded_file.read())
                tmp_file_path = tmp_file.name
            
            # Load PDF
            loader = PyPDFLoader(tmp_file_path)
            docs = loader.load()
            
            # Add file name to metadata
            for doc in docs:
                doc.metadata['source_file'] = uploaded_file.name
            
            documents.extend(docs)
            
            # Clean up temporary file
            os.unlink(tmp_file_path)
            
            st.success(f"✅ Processed: {uploaded_file.name} ({len(docs)} pages)")
            
        except Exception as e:
            st.error(f"❌ Error processing {uploaded_file.name}: {e}")
    
    if not documents:
        return None, []
    
    # Split documents into chunks
    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=1000,
        chunk_overlap=200,
        length_function=len,
        separators=["\n\n", "\n", " ", ""]
    )
    
    text_chunks = text_splitter.split_documents(documents)
    
    # Add metadata to chunks
    for i, text in enumerate(text_chunks):
        text.metadata.update({
            "chunk_id": i,
            "chunk_size": len(text.page_content)
        })
    
    st.info(f"✂️ Created {len(text_chunks)} text chunks")
    
    # Create FAISS vector store
    try:
        vectorstore = FAISS.from_documents(text_chunks, embeddings)
        st.success(f"✅ Successfully created vector database with {len(text_chunks)} chunks!")
        return vectorstore, text_chunks
    except Exception as e:
        st.error(f"❌ Error creating vector database: {e}")
        return None, []

def create_qa_chain(llm, vectorstore, k=5):
    """Create a question-answering chain with retrieval"""
    if not vectorstore or not llm:
        return None
    
    prompt_template = """Use the following context to answer the question. If you cannot find the answer in the context, say "I cannot find this information in the provided documents."

Context: {context}

Question: {question}

Answer:"""

    PROMPT = PromptTemplate(
        template=prompt_template,
        input_variables=["context", "question"]
    )
    
    try:
        qa_chain = RetrievalQA.from_chain_type(
            llm=llm,
            chain_type="stuff",
            retriever=vectorstore.as_retriever(search_kwargs={"k": k}),
            chain_type_kwargs={"prompt": PROMPT},
            return_source_documents=True
        )
        return qa_chain
    except Exception as e:
        st.error(f"Error creating QA chain: {e}")
        return None

def ask_question(qa_chain, question):
    """Ask a question and get an answer with sources"""
    if not qa_chain:
        return None
    
    try:
        result = qa_chain({"query": question})
        
        response = {
            "question": question,
            "answer": result["result"],
            "source_documents": result.get("source_documents", [])
        }
        
        return response
        
    except Exception as e:
        st.error(f"❌ Error processing question: {e}")
        return None

def search_similar_chunks(vectorstore, query, k=5):
    """Search for similar chunks without generating an answer"""
    if not vectorstore:
        return []
    
    try:
        results = vectorstore.similarity_search(query, k=k)
        return results
    except Exception as e:
        st.error(f"Error searching: {e}")
        return []

# Main App Interface
def main():
    st.markdown('<h1 class="main-header">📚 PDF RAG System</h1>', unsafe_allow_html=True)
    st.markdown("Upload PDF documents and ask questions about their content using AI-powered retrieval!")
    
    # Sidebar for configuration
    with st.sidebar:
        st.markdown('<h2 class="sidebar-header">⚙️ Configuration</h2>', unsafe_allow_html=True)
        
        # Model configuration
        st.subheader("🤖 Model Settings")
        llm_model = st.selectbox(
            "Language Model",
            ["google/flan-t5-small", "google/flan-t5-base"],
            help="Choose the language model (smaller models are faster)"
        )
        
        embedding_model = st.selectbox(
            "Embedding Model",
            ["all-MiniLM-L6-v2", "sentence-transformers/all-mpnet-base-v2"],
            help="Choose the embedding model"
        )
        
        retrieval_k = st.slider(
            "Number of chunks to retrieve",
            min_value=1,
            max_value=10,
            value=5,
            help="How many relevant chunks to use for answering questions"
        )
        
        st.subheader("💾 Vector Store")
        st.info("Using FAISS (local vector storage)")
        
        # Option to save/load vector store
        if st.session_state.vectorstore:
            if st.button("💾 Save Vector Store"):
                try:
                    # Save vector store to session state or file
                    st.session_state.vectorstore.save_local("faiss_index")
                    st.success("Vector store saved!")
                except Exception as e:
                    st.error(f"Error saving: {e}")
    
    # Main content area
    col1, col2 = st.columns([1, 1])
    
    with col1:
        st.subheader("📁 Upload Documents")
        uploaded_files = st.file_uploader(
            "Choose PDF files",
            type=['pdf'],
            accept_multiple_files=True,
            help="Upload one or more PDF files to analyze"
        )
        
        if st.button("🚀 Process Documents", type="primary"):
            if not uploaded_files:
                st.warning("Please upload at least one PDF file.")
            else:
                with st.spinner("Processing documents..."):
                    # Setup models
                    llm = setup_llm(llm_model)
                    embeddings = setup_embeddings(embedding_model)
                    
                    if llm and embeddings:
                        # Process files
                        vectorstore, text_chunks = process_uploaded_files(uploaded_files, embeddings)
                        
                        if vectorstore:
                            # Create QA chain
                            qa_chain = create_qa_chain(llm, vectorstore, k=retrieval_k)
                            
                            if qa_chain:
                                # Store in session state
                                st.session_state.qa_chain = qa_chain
                                st.session_state.vectorstore = vectorstore
                                st.session_state.documents_processed = True
                                
                                st.balloons()
                                st.success("🎉 Documents processed successfully! You can now ask questions.")
                            else:
                                st.error("Failed to create QA chain.")
                    else:
                        st.error("Failed to load models.")
    
    with col2:
        st.subheader("💬 Ask Questions")
        
        if st.session_state.documents_processed:
            question = st.text_input(
                "Your question:",
                placeholder="What are the main topics discussed in the documents?",
                help="Ask any question about your uploaded documents"
            )
            
            col2a, col2b = st.columns([1, 1])
            
            with col2a:
                if st.button("🔍 Get Answer"):
                    if question:
                        with st.spinner("Searching for answer..."):
                            result = ask_question(st.session_state.qa_chain, question)
                            
                            if result:
                                # Add to chat history
                                st.session_state.chat_history.append({
                                    "question": question,
                                    "answer": result["answer"],
                                    "sources": result["source_documents"]
                                })
                                
                                # Display answer
                                st.subheader("💡 Answer:")
                                st.write(result["answer"])
                                
                                # Display sources
                                if result["source_documents"]:
                                    st.subheader("📚 Sources:")
                                    for i, doc in enumerate(result["source_documents"][:3]):
                                        with st.expander(f"Source {i+1}: {doc.metadata.get('source_file', 'Unknown')}"):
                                            st.write(doc.page_content[:500] + "..." if len(doc.page_content) > 500 else doc.page_content)
                    else:
                        st.warning("Please enter a question.")
            
            with col2b:
                if st.button("🔍 Search Similar"):
                    if question:
                        with st.spinner("Searching for similar content..."):
                            results = search_similar_chunks(st.session_state.vectorstore, question, k=5)
                            
                            if results:
                                st.subheader("🔍 Similar Content:")
                                for i, doc in enumerate(results):
                                    with st.expander(f"Match {i+1}: {doc.metadata.get('source_file', 'Unknown')}"):
                                        st.write(doc.page_content[:300] + "..." if len(doc.page_content) > 300 else doc.page_content)
        else:
            st.info("👆 Please upload and process documents first to start asking questions.")
    
    # Chat History
    if st.session_state.chat_history:
        st.subheader("📝 Chat History")
        
        for i, chat in enumerate(reversed(st.session_state.chat_history[-5:])):  # Show last 5
            with st.expander(f"Q: {chat['question'][:50]}..."):
                st.write("**Question:**", chat['question'])
                st.write("**Answer:**", chat['answer'])
                
                if chat['sources']:
                    st.write("**Sources:**")
                    for j, doc in enumerate(chat['sources'][:2]):  # Show top 2 sources
                        st.write(f"{j+1}. {doc.metadata.get('source_file', 'Unknown')}")
    
    # Clear session button
    if st.session_state.documents_processed:
        if st.button("🗑️ Clear Session"):
            st.session_state.qa_chain = None
            st.session_state.vectorstore = None
            st.session_state.documents_processed = False
            st.session_state.chat_history = []
            st.success("Session cleared! You can upload new documents.")
            st.rerun()

if __name__ == "__main__":
    main()