app.py

import os
import streamlit as st
from langchain_chroma import Chroma
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain.chains.question_answering import load_qa_chain
from langchain.memory import ConversationBufferMemory
from langchain_core.prompts import PromptTemplate
from langchain_groq import ChatGroq
from dotenv import load_dotenv
from sentence_transformers import SentenceTransformer

st.title("HocamBot")

# Load environment variables
load_dotenv()
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
assert GROQ_API_KEY, "GROQ_API_KEY environment variable not set."

# One-time setup in session state
if 'initialized' not in st.session_state:
    st.session_state.initialized = False

    try:
        with st.spinner("Initializing..."):
            # Initialize embeddings model
            model_path = "sentence-transformers/all-MiniLM-L12-v2"  # Use a smaller, faster model
            
            st.session_state.embedding_function = HuggingFaceEmbeddings(
                model_name=model_path,
                model_kwargs={'device': 'cpu'},
                encode_kwargs={'normalize_embeddings': False}
            )
            
            # Set up document search
            persist_directory = "doc_db"
            st.session_state.docsearch = Chroma(
                persist_directory=persist_directory,
                embedding_function=st.session_state.embedding_function
            )

            # Initialize ChatGroq model
            st.session_state.chat_model = ChatGroq(
                model="llama-3.1-8b-instant",
                temperature=0,
                api_key=GROQ_API_KEY
            )

            # Define prompt template and memory
            template = """You are a chatbot having a conversation with a human. Your name is Devrim.
            Given the following extracted parts of a long document and a question, create a final answer. If the answer is not in the document or irrelevant, just say that you don't know, don't try to make up an answer.
            {context}
            {chat_history}
            Human: {human_input}
            Chatbot:"""

            prompt = PromptTemplate(
                input_variables=["chat_history", "human_input", "context"], template=template
            )
            st.session_state.memory = ConversationBufferMemory(memory_key="chat_history", input_key="human_input")

            # Load QA chain
            st.session_state.qa_chain = load_qa_chain(
                llm=st.session_state.chat_model, 
                chain_type="stuff", 
                memory=st.session_state.memory, 
                prompt=prompt
            )

            st.session_state.initialized = True
            st.success("Initialization successful.")

    except Exception as e:
        st.session_state.initialized = False
        st.error(f"Initialization failed: {e}")

# Clear chat history buttons
if st.button("Clear Chat History"):
    if 'memory' in st.session_state:
        st.session_state.memory.clear()
    st.rerun()  # Refresh the app to reflect the cleared history

# Display chat history if initialized
if st.session_state.initialized and 'memory' in st.session_state:
    if st.session_state.memory.buffer_as_messages:
        for message in st.session_state.memory.buffer_as_messages:
            if message.type == "ai":
                st.chat_message(name="ai", avatar="🤖").write(message.content)
            else:
                st.chat_message(name="human", avatar="👤").write(message.content)

# Input for new query
query = st.chat_input("Ask something")
if query:
    try:
        with st.spinner("Answering..."):
            # Perform similarity search and get response
            docs = st.session_state.docsearch.similarity_search(query, k=1)  # Reduced k for speed
            response = st.session_state.qa_chain(
                {"input_documents": docs, "human_input": query}, 
                return_only_outputs=True
            )["output_text"]

            # Display new message
            st.chat_message(name="human", avatar="👤").write(query)
            st.chat_message(name="ai", avatar="🤖").write(response)

    except Exception as e:
        st.error(f"An error occurred: {e}")