app.py

import gradio as gr
import re
import ast
import pandas as pd
from geopy.geocoders import Nominatim
from geopy.distance import geodesic
from geopy.exc import GeocoderTimedOut
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
import google.generativeai as genai
from dotenv import load_dotenv
from supabase import create_client
import os

# --- Supabase & Gemini ---
load_dotenv()
SUPABASE_URL = os.getenv("SUPABASE_URL")
SUPABASE_KEY = os.getenv("SBASEKEY")
API_KEY = os.getenv("GEMINI_API_KEY")

supabase = create_client(SUPABASE_URL, SUPABASE_KEY)
genai.configure(api_key=API_KEY)

# --- Load data dari Supabase ---
def fetch_data_from_supabase():
    response = supabase.table("Maps").select("*").execute()
    return pd.DataFrame(response.data)

df = fetch_data_from_supabase()

# --- Ekstraksi Kata Kunci ---
def extract_keywords(user_input):
    prompt = f"""
    Ekstrak 3–7 kata kunci penting dari deskripsi wisata berikut:
    "{user_input}"
    Tulis langsung sebagai list Python tanpa variabel apapun.
    """
    try:
        response = genai.GenerativeModel("gemini-1.5-flash").generate_content(prompt)
        matches = re.findall(r'\[.*?\]', response.text)
        if matches:
            return ast.literal_eval(matches[0])
        else:
            return []
    except Exception as e:
        return [f"Error: {e}"]

# --- Lokasi ---
def get_coordinates_from_location(location_name):
    try:
        geolocator = Nominatim(user_agent="geoapi")
        location = geolocator.geocode(location_name, timeout=10)
        return (location.latitude, location.longitude) if location else (None, None)
    except GeocoderTimedOut:
        return (None, None)

def get_location_name_from_coordinates(lat, lon):
    try:
        geolocator = Nominatim(user_agent="geoapi")
        location = geolocator.reverse((lat, lon), timeout=10)
        return location.address if location else "Tidak ditemukan"
    except GeocoderTimedOut:
        return "Tidak ditemukan"

# --- Enhancement untuk deskripsi singkat ---
def enhance_description_with_gemini(short_desc):
    prompt = f"""
    Deskripsi berikut terlalu singkat: "{short_desc}"
    Tolong kembangkan menjadi paragraf singkat (2–3 kalimat) yang menggambarkan keinginan wisata pengguna secara lebih lengkap.
    Contohnya: sebutkan suasana, aktivitas, atau lokasi ideal.
    """
    try:
        response = genai.GenerativeModel("gemini-1.5-flash").generate_content(prompt)
        return response.text.strip()
    except Exception as e:
        return short_desc

# --- TF-IDF dan Cosine Similarity ---
def prepare_and_recommend(df, user_description):
    tfidf = TfidfVectorizer()
    tfidf_matrix = tfidf.fit_transform(df['deskripsi'].astype(str).tolist() + [user_description])
    similarity = cosine_similarity(tfidf_matrix[-1], tfidf_matrix[:-1]).flatten()
    df['similarity'] = similarity
    return df.sort_values(by='similarity', ascending=False).head(10)

# --- Jarak Lokasi ---
def sort_by_nearest_location(df, user_lat, user_lon):
    df['distance_km'] = df.apply(
        lambda row: geodesic((user_lat, user_lon), (row['latitude'], row['longitude'])).km,
        axis=1
    )
    df['distance_km'] = df['distance_km'].round(2)
    return df.sort_values(by='distance_km')

# --- Fungsi Utama ---
def wisata_rekomendasi(deskripsi, lokasi):
    if df.empty:
        return "Data tidak tersedia.", pd.DataFrame([["Data tidak tersedia"]], columns=["nama"])

    if len(deskripsi.strip().split()) < 3 or len(deskripsi.strip()) < 20:
        deskripsi = enhance_description_with_gemini(deskripsi)

    # Lokasi → Koordinat
    lat, lon = get_coordinates_from_location(lokasi)
    if lat is None or lon is None:
        return "Lokasi tidak ditemukan.", pd.DataFrame([["Lokasi tidak ditemukan"]], columns=["nama"])

    # Tambahkan lokasi ke deskripsi
    deskripsi_lengkap = f"{deskripsi} di sekitar {lokasi}"

    keywords = extract_keywords(deskripsi_lengkap)
    if "Error:" in str(keywords):
        return f"Kata kunci gagal diambil: {keywords[0]}", pd.DataFrame([[keywords[0]]], columns=["nama"])

    user_description_joined = " ".join(keywords)

    top_place = prepare_and_recommend(df.copy(), user_description_joined)
    top_place = top_place[top_place['total_ulasan'] > 10]
    sorted_place = sort_by_nearest_location(top_place, lat, lon)

    sorted_place = sorted_place[sorted_place["gambar"].apply(lambda x: isinstance(x, str) and x.startswith("https"))]

    # Urutkan berdasarkan similarity tertinggi dan tampilkan kolom similarity
    sorted_place = sorted_place.sort_values(by='similarity', ascending=False)

    return f"Kata kunci: {', '.join(keywords)}", sorted_place[[
        "id", "nama", "alamat", "distance_km", "deskripsi", "harga", "rating", "total_ulasan", "gambar", "similarity"
    ]]

# --- Gradio UI ---
demo = gr.Interface(
    fn=wisata_rekomendasi,
    inputs=[
        gr.Textbox(label="Deskripsi Wisata yang Anda Inginkan"),
        gr.Textbox(label="Lokasi Anda (Contoh: Cilacap, Jawa Tengah, Indonesia)"),
    ],
    outputs=[
        gr.Textbox(label="Kata Kunci yang Diekstrak"),
        gr.Dataframe(
            headers=["id", "nama", "alamat", "distance_km", "deskripsi", "harga", "rating", "total_ulasan", "gambar", "similarity"],
            label="Rekomendasi Tempat Wisata"
        )
    ],
    title="Sistem Rekomendasi Wisata",
    description="Masukkan deskripsi dan lokasi, lalu dapatkan rekomendasi tempat wisata terdekat beserta skor kecocokannya."
)

demo.launch()