app.py

import streamlit as st
import os
from dotenv import load_dotenv 
from langsmith import traceable

from app.chat import initialize_session_state, display_chat_history
from app.data_loader import get_data, load_docs
from app.document_processor import process_documents, save_vector_store, load_vector_store
from app.prompts import sahabat_prompt
from langchain_community.llms import Replicate
from langchain.memory import ConversationBufferMemory
from langchain.chains import ConversationalRetrievalChain
from langchain_community.document_transformers import LongContextReorder

load_dotenv()

VECTOR_STORE_PATH = "vector_store_data"
DATA_DIR = "data"

@traceable(name="Create RAG Conversational Chain")
def create_conversational_chain(vector_store):
    llm = Replicate(
        model="fauziisyrinapridal/sahabat-ai-v1:afb9fa89fe786362f619fd4fef34bd1f7a4a4da23073d8a6fbf54dcbe458f216",
        model_kwargs={"temperature": 0.1, "top_p": 0.9, "max_new_tokens": 6000}
    )

    memory = ConversationBufferMemory(
        memory_key="chat_history",
        return_messages=True,
        output_key='answer'
    )

    chain = ConversationalRetrievalChain.from_llm(
        llm,
        retriever=vector_store.as_retriever(search_kwargs={"k": 6}),
        combine_docs_chain_kwargs={"prompt": sahabat_prompt},
        return_source_documents=True,
        memory=memory
    )

    return chain

def reorder_embedding(docs):
    reordering = LongContextReorder()
    return reordering.transform_documents(docs)

def get_latest_data_timestamp(folder):
    latest_time = 0
    for root, _, files in os.walk(folder):
        for file in files:
            path = os.path.join(root, file)
            file_time = os.path.getmtime(path)
            latest_time = max(latest_time, file_time)
    return latest_time

def vector_store_is_outdated():
    if not os.path.exists(VECTOR_STORE_PATH):
        return True
    vector_store_time = os.path.getmtime(VECTOR_STORE_PATH)
    data_time = get_latest_data_timestamp(DATA_DIR)
    return data_time > vector_store_time

@traceable(name="Main Chatbot RAG App")
def main():
    initialize_session_state()
    get_data()

    if len(st.session_state['history']) == 0:
        if vector_store_is_outdated():
            docs = load_docs()
            if len(docs) > 0:
                reordered_docs = reorder_embedding(docs)
                vector_store = process_documents(reordered_docs)
                save_vector_store(vector_store)
            else:
                st.warning("Tidak ada dokumen ditemukan di folder 'data/'. Chatbot tetap bisa digunakan, tapi tanpa konteks dokumen.")
                vector_store = None
    else:
        vector_store = load_vector_store()

    st.session_state['vector_store'] = vector_store


    if st.session_state['vector_store'] is not None:
        chain = create_conversational_chain(st.session_state['vector_store'])
        display_chat_history(chain)

if __name__ == "__main__":
    main()