# Import Library yang Diperlukan
import gradio as gr
import gspread
from oauth2client.service_account import ServiceAccountCredentials
from llama_cpp import Llama
from llama_index.core import VectorStoreIndex, Settings
from llama_index.core.node_parser import SentenceSplitter
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.llms.llama_cpp import LlamaCPP
from huggingface_hub import hf_hub_download
from llama_index.core.llms import ChatMessage
from llama_index.core.chat_engine.condense_plus_context import CondensePlusContextChatEngine

# ===================================
# 1️⃣ Fungsi untuk Membaca Google Spreadsheet
# ===================================
def read_google_sheet():
    # Tentukan scope akses ke Google Sheets & Drive
    scope = ["https://spreadsheets.google.com/feeds", "https://www.googleapis.com/auth/drive"]
    
    # Load kredensial dari file credentials.json
    creds = ServiceAccountCredentials.from_json_keyfile_name("credentials.json", scope)
    client = gspread.authorize(creds)
    
    # 📌 GANTI BAGIAN INI SESUAI SPREADSHEET ANDA
    spreadsheet = client.open("datatarget")  # 🔹 Ganti dengan nama spreadsheet Anda
    sheet = spreadsheet.datatarget  # 🔹 Jika ingin sheet lain, ganti dengan spreadsheet.worksheet("NamaSheet")

    # Ambil semua data dalam bentuk list (baris & kolom)
    data = sheet.get_all_values()

    # Format ulang data menjadi satu teks panjang (dapat disesuaikan)
    formatted_text = "\n".join([" | ".join(row) for row in data])
    
    return formatted_text

# ===================================
# 2️⃣ Fungsi untuk Mengunduh Model Llama
# ===================================
def initialize_llama_model():
    model_path = hf_hub_download(
        repo_id="TheBLoke/zephyr-7b-beta-GGUF",  # 📌 Repo model HuggingFace
        filename="zephyr-7b-beta.Q4_K_M.gguf",  # 📌 Nama file model
        cache_dir="./models"
    )
    return model_path

# ===================================
# 3️⃣ Inisialisasi Model dan Pengaturan
# ===================================
def initialize_settings(model_path):
    Settings.llm = LlamaCPP(
        model_path=model_path,
        temperature=0.7,
    )

# ===================================
# 4️⃣ Inisialisasi Index dari Data Spreadsheet
# ===================================
def initialize_index():
    # 🔹 Ambil teks dari Google Spreadsheet
    text_data = read_google_sheet()

    # 🔹 Konversi teks ke dalam format dokumen
    documents = [text_data]

    # 🔹 Proses data menjadi node untuk vektor embedding
    parser = SentenceSplitter(chunk_size=150, chunk_overlap=10)
    nodes = parser.get_nodes_from_documents(documents)
    
    # 🔹 Gunakan model embedding
    embedding = HuggingFaceEmbedding("sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")
    Settings.embed_model = embedding
    
    # 🔹 Buat index vektor
    index = VectorStoreIndex(nodes)
    return index

# ===================================
# 5️⃣ Inisialisasi Mesin Chatbot
# ===================================
def initialize_chat_engine(index):
    retriever = index.as_retriever(similarity_top_k=3)
    chat_engine = CondensePlusContextChatEngine.from_defaults(
        retriever=retriever,
        verbose=True,
    )
    return chat_engine

# ===================================
# 6️⃣ Fungsi untuk Menghasilkan Respons Chatbot
# ===================================
def generate_response(message, history, chat_engine):
    if history is None:
        history = []
        
    chat_messages = [
        ChatMessage(
            role="system",
            content="Anda adalah chatbot yang menjawab dalam bahasa Indonesia berdasarkan dokumen di Google Spreadsheet."
        ),
    ]
    
    response = chat_engine.stream_chat(message)
    text = "".join(response.response_gen)  # 🔹 Gabungkan semua token menjadi string
    
    history.append((message, text))
    return history

# ===================================
# 7️⃣ Fungsi Utama untuk Menjalankan Aplikasi
# ===================================
def main():
    # 🔹 Unduh model dan inisialisasi pengaturan
    model_path = initialize_llama_model()
    initialize_settings(model_path)  

    # 🔹 Inisialisasi index dan chat engine
    index = initialize_index()
    chat_engine = initialize_chat_engine(index)  

    # 🔹 Fungsi untuk chat
    def chatbot_response(message, history):
        return generate_response(message, history, chat_engine)  

    # 🔹 Luncurkan Gradio UI
    gr.Interface(
        fn=chatbot_response,
        inputs=["text"],
        outputs=["text"],
    ).launch()

if __name__ == "__main__":
    main()