app.py

import os
import chromadb
import requests
import streamlit as st
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
from langchain_groq import ChatGroq
from langchain.document_loaders import PDFPlumberLoader
from langchain_experimental.text_splitter import SemanticChunker
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_chroma import Chroma
from prompts import rag_prompt, relevancy_prompt, relevant_context_picker_prompt, response_synth

# Set API Keys
os.environ["GROQ_API_KEY"] = st.secrets.get("GROQ_API_KEY", "")

# Load LLM models
llm_judge = ChatGroq(model="deepseek-r1-distill-llama-70b")
rag_llm = ChatGroq(model="mixtral-8x7b-32768")

llm_judge.verbose = True
rag_llm.verbose = True

# Clear ChromaDB cache to fix tenant issue
chromadb.api.client.SharedSystemClient.clear_system_cache()

st.title("🔍 PDF-based RAG System")

# Initialize session state variables
if "vector_store" not in st.session_state:
    st.session_state.vector_store = None
if "documents" not in st.session_state:
    st.session_state.documents = None
if "pdf_path" not in st.session_state:
    st.session_state.pdf_path = None  
if "pdf_loaded" not in st.session_state:
    st.session_state.pdf_loaded = False
if "chunked" not in st.session_state:
    st.session_state.chunked = False
if "vector_created" not in st.session_state:
    st.session_state.vector_created = False

# Step 1: Choose PDF Source
pdf_source = st.radio("Upload or provide a link to a PDF:", ["Upload a PDF file", "Enter a PDF URL"], index=0, horizontal=True)

if pdf_source == "Upload a PDF file":
    uploaded_file = st.file_uploader("Upload your PDF file", type="pdf")
    if uploaded_file:
        st.session_state.pdf_path = "temp.pdf"
        with open(st.session_state.pdf_path, "wb") as f:
            f.write(uploaded_file.getbuffer())
        st.session_state.pdf_loaded = False
        st.session_state.chunked = False
        st.session_state.vector_created = False
        st.success("✅ PDF Uploaded Successfully!")

elif pdf_source == "Enter a PDF URL":
    pdf_url = st.text_input("Enter PDF URL:", value="https://arxiv.org/pdf/2406.06998")
    if pdf_url and st.session_state.pdf_path is None:
        with st.spinner("Downloading PDF..."):
            try:
                response = requests.get(pdf_url)
                if response.status_code == 200:
                    st.session_state.pdf_path = "temp.pdf"
                    with open(st.session_state.pdf_path, "wb") as f:
                        f.write(response.content)
                    st.session_state.pdf_loaded = False
                    st.session_state.chunked = False
                    st.session_state.vector_created = False
                    st.success("✅ PDF Downloaded Successfully!")
                else:
                    st.error("❌ Failed to download PDF. Check the URL.")
            except Exception as e:
                st.error(f"Error downloading PDF: {e}")

# Step 2: Process PDF
if st.session_state.pdf_path and not st.session_state.pdf_loaded:
    with st.spinner("Loading and processing PDF..."):
        loader = PDFPlumberLoader(st.session_state.pdf_path)
        docs = loader.load()
        st.session_state.documents = docs
        st.session_state.pdf_loaded = True
        st.success(f"✅ **PDF Loaded!** Total Pages: {len(docs)}")

# Step 3: Chunking (Only if Not Already Done)
if st.session_state.pdf_loaded and not st.session_state.chunked:
    with st.spinner("Chunking the document..."):
        model_name = "nomic-ai/modernbert-embed-base"
        embedding_model = HuggingFaceEmbeddings(model_name=model_name, model_kwargs={'device': 'cpu'}, encode_kwargs={'normalize_embeddings': False})
        text_splitter = SemanticChunker(embedding_model)
        documents = text_splitter.split_documents(st.session_state.documents)
        st.session_state.documents = documents
        st.session_state.chunked = True
        st.success(f"✅ **Document Chunked!** Total Chunks: {len(documents)}")

# Step 4: Setup Vectorstore
if st.session_state.chunked and not st.session_state.vector_created:
    with st.spinner("Creating vector store..."):
        vector_store = Chroma(
            collection_name="deepseek_collection",
            collection_metadata={"hnsw:space": "cosine"},
            embedding_function=embedding_model
        )
        vector_store.add_documents(st.session_state.documents)
        num_documents = len(vector_store.get()["documents"])
        st.session_state.vector_store = vector_store
        st.session_state.vector_created = True
        st.success(f"✅ **Vector Store Created!** Total documents stored: {num_documents}")

# Step 5: Query Input
if st.session_state.vector_created:
    query = st.text_input("🔍 Enter a Query:")
    if query:
        with st.spinner("Retrieving relevant contexts..."):
            retriever = st.session_state.vector_store.as_retriever(search_type="similarity", search_kwargs={"k": 5})
            contexts = retriever.invoke(query)
            context_texts = [doc.page_content for doc in contexts]

        st.success(f"✅ **Retrieved {len(context_texts)} Contexts!**")
        for i, text in enumerate(context_texts, 1):
            st.write(f"**Context {i}:** {text[:500]}...")

        # Step 6: Context Relevancy Checker
        with st.spinner("Evaluating context relevancy..."):
            context_relevancy_checker_prompt = PromptTemplate(input_variables=["retriever_query", "context"], template=relevancy_prompt)
            context_relevancy_chain = LLMChain(llm=llm_judge, prompt=context_relevancy_checker_prompt, output_key="relevancy_response")
            relevancy_response = context_relevancy_chain.invoke({"context": context_texts, "retriever_query": query})

        st.subheader("🟥 Context Relevancy Evaluation")
        st.json(relevancy_response['relevancy_response'])

        # Step 7: Selecting Relevant Contexts
        with st.spinner("Selecting the most relevant contexts..."):
            relevant_prompt = PromptTemplate(input_variables=["relevancy_response"], template=relevant_context_picker_prompt)
            pick_relevant_context_chain = LLMChain(llm=llm_judge, prompt=relevant_prompt, output_key="context_number")
            relevant_response = pick_relevant_context_chain.invoke({"relevancy_response": relevancy_response['relevancy_response']})

        st.subheader("🟦 Pick Relevant Context Chain")
        st.json(relevant_response['context_number'])

        # Step 8: Retrieving Context for Response Generation
        with st.spinner("Retrieving final context..."):
            context_prompt = PromptTemplate(input_variables=["context_number", "context"], template=response_synth)
            relevant_contexts_chain = LLMChain(llm=llm_judge, prompt=context_prompt, output_key="relevant_contexts")
            final_contexts = relevant_contexts_chain.invoke({"context_number": relevant_response['context_number'], "context": context_texts})

        st.subheader("🟥 Relevant Contexts Extracted")
        st.json(final_contexts['relevant_contexts'])

        # Step 9: Generate Final Response
        with st.spinner("Generating the final answer..."):
            final_prompt = PromptTemplate(input_variables=["query", "context"], template=rag_prompt)
            response_chain = LLMChain(llm=rag_llm, prompt=final_prompt, output_key="final_response")
            final_response = response_chain.invoke({"query": query, "context": final_contexts['relevant_contexts']})

        st.subheader("🟥 RAG Final Response")
        st.success(final_response['final_response'])

else:
    st.warning("📄 Please upload or provide a PDF URL first.")