src/streamlit_app.py

import os
import re
import logging
from uuid import uuid4
from pathlib import Path
from dotenv import load_dotenv
import streamlit as st

from langchain.chains import create_history_aware_retriever, create_retrieval_chain
from langchain.chains.combine_documents import create_stuff_documents_chain
from langchain_community.chat_message_histories import ChatMessageHistory
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_groq import ChatGroq
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.document_loaders import PyPDFLoader
from langchain_chroma import Chroma
import torch

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

# Set up proper cache directories
def setup_environment():
    cache_dir = Path("/tmp/cache")
    cache_dir.mkdir(exist_ok=True)
    os.environ['HF_HOME'] = str(cache_dir / "huggingface")
    os.environ['STREAMLIT_HOME'] = str(cache_dir / "streamlit")

setup_environment()

# Load environment variables
load_dotenv()
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
PDF_PATH = os.getenv("PDF_PATH", "nivakaran.pdf")  # Changed to direct filename

# Validate environment variables
if not all([GROQ_API_KEY]):
    st.error("Missing required environment variables")
    st.stop()

# Verify PDF exists
if not Path(PDF_PATH).exists():
    st.error(f"PDF file not found at: {PDF_PATH}")
    st.stop()

# Initialize RAG components with proper device handling
try:
    # Force CPU and disable metal for sentence-transformers
    os.environ['PYTORCH_ENABLE_MPS_FALLBACK'] = '1'
    os.environ['PYTORCH_MPS_HIGH_WATERMARK_RATIO'] = '0.0'
    
    embeddings = HuggingFaceEmbeddings(
        model_name="sentence-transformers/all-MiniLM-L6-v2",
        model_kwargs={'device': 'cpu'},
        encode_kwargs={'normalize_embeddings': True}
    )
except Exception as e:
    logger.error(f"Failed to initialize embeddings: {str(e)}")
    st.error("Failed to initialize embeddings. Please try again later.")
    st.stop()

llm = ChatGroq(model_name="Deepseek-R1-Distill-Llama-70b", temperature=0.1)

# Process PDF into vectorstore
def process_pdf(file_path: str):
    try:
        loader = PyPDFLoader(file_path)
        documents = loader.load()
        text_splitter = RecursiveCharacterTextSplitter(chunk_size=5000, chunk_overlap=500)
        splits = text_splitter.split_documents(documents)
        
        vectorstore = Chroma.from_documents(
            documents=splits,
            embedding=embeddings,
            persist_directory="/tmp/chroma_db"
        )
        logger.info(f"PDF {file_path} processed successfully")
        return vectorstore
    except Exception as e:
        logger.error(f"Failed to process PDF: {str(e)}")
        st.error("PDF processing failed")
        st.stop()

# Initialize vectorstore and retriever
try:
    vectorstore = process_pdf(PDF_PATH)
    retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
except Exception as e:
    logger.error(f"Failed to initialize vectorstore: {str(e)}")
    st.error("Failed to initialize document store. Please try again later.")
    st.stop()

# [Rest of your existing Streamlit UI code remains the same...]

# System prompt for the assistant
system_prompt = """You are Max, a friendly and professional chatbot designed to 
assist visitors to Nivakaran's portfolio website. Your primary goal 
is to provide accurate, clear, and helpful information about Nivakaran, based 
on the following context:

{context}

Your responses should be:
1. Informative and relevant, directly addressing the visitor's questions about Nivakaran's skills, 
projects, experience, and background.
2. Concise but thorough enough to give visitors a clear understanding of Nivakaran's expertise.
3. Engaging and approachable, maintaining a professional yet conversational tone.
4. Honest about what is available in the provided context; if you don't know an answer, politely 
say so and suggest the visitor explore other sections of the portfolio or contact Nivakaran directly.
5. Focused on helping visitors understand Nivakaran's capabilities and what makes him stand out 
as a developer and professional.
6. Ready to provide examples, explanations, or links to portfolio projects when relevant.

Avoid providing generic or unrelated information. Always tailor your answers to 
highlight Nivakaran's strengths and the unique value he brings.
"""

# Streamlit app UI
st.set_page_config(page_title="Nivakaran's Portfolio Assistant", page_icon="💬")
st.title("💬 Nivakaran's Portfolio Assistant")

# Session ID and message history
if "session_id" not in st.session_state:
    st.session_state.session_id = str(uuid4())
if "history" not in st.session_state:
    st.session_state.history = ChatMessageHistory()

# Display chat history
for message in st.session_state.history.messages:
    role = "user" if message.type == "human" else "assistant"
    with st.chat_message(role):
        st.markdown(message.content)

# User input
if user_input := st.chat_input("Ask me something about Nivakaran..."):
    with st.chat_message("user"):
        st.markdown(user_input)
    st.session_state.history.add_user_message(user_input)

    try:
        last_messages = st.session_state.history.messages[-6:]

        # Contextualize question based on history
        contextualize_q_prompt = ChatPromptTemplate.from_messages([
            ("system", "Given a chat history and the latest user question which might reference context in the chat history, formulate a standalone question which can be understood without the chat history. Return just the question and nothing else."),
            MessagesPlaceholder("chat_history"),
            ("human", "{input}")
        ])
        
        history_aware_retriever = create_history_aware_retriever(
            llm, retriever, contextualize_q_prompt
        )

        # RAG chain
        qa_prompt = ChatPromptTemplate.from_messages([
            ("system", system_prompt),
            MessagesPlaceholder("chat_history"),
            ("human", "{input}")
        ])
        
        question_answer_chain = create_stuff_documents_chain(llm, qa_prompt)
        rag_chain = create_retrieval_chain(history_aware_retriever, question_answer_chain)

        result = rag_chain.invoke({
            "input": user_input,
            "chat_history": last_messages
        })
        
        raw_answer = result["answer"]
        # Clean out <think>...</think> junk and any other unwanted artifacts
        cleaned_answer = re.sub(r"<think>.*?</think>\s*", "", raw_answer, flags=re.DOTALL).strip()
        cleaned_answer = re.sub(r"<\|.*?\|>", "", cleaned_answer).strip()

        with st.chat_message("assistant"):
            st.markdown(cleaned_answer)

        st.session_state.history.add_ai_message(cleaned_answer)

    except Exception as e:
        logger.error(f"Error during RAG processing: {str(e)}")
        st.error("Sorry, I encountered an error while processing your request. Please try again.")