Update api.py

api.py

@@ -6,37 +6,54 @@ import os
 import re
 import requests
 from bs4 import BeautifulSoup
+from tensorflow.keras.models import load_model 
 
 app = FastAPI(
-    title="API Deteksi Hoax Naive Bayes",
-    description="API untuk mendeteksi berita hoax dari teks atau URL portal berita.",
+    title="API Deteksi Hoax Multi-Model",
+    description="API untuk mendeteksi berita hoax menggunakan pilihan model.",
     version="1.0.0"
 )
 
-# --- KONFIGURASI CORS ---
 app.add_middleware(
     CORSMiddleware,
     allow_origins=["https://deteksi-berita-hoax-kappa.vercel.app/"], 
     allow_credentials=True,
-    allow_methods=["https://deteksi-berita-hoax-kappa.vercel.app/"],
-    allow_headers=["https://deteksi-berita-hoax-kappa.vercel.app/"],
+    allow_methods=["*"],
+    allow_headers=["*"],
 )
-# Load Model saat server menyala
-#MODEL_PATH = 'model_hoax_complete.pkl'
-MODEL_PATH = 'lstm_fake_news_model.h5'
+
+# --- 1. LOAD KEDUA MODEL ---
+models = {
+    "naive_bayes": None,
+    "lstm": None
+}
+
+# Load Model Naive Bayes
+PATH_NB = 'model_hoax_complete.pkl'
 try:
-    if os.path.exists(MODEL_PATH):
-        model = joblib.load(MODEL_PATH)
-        print("Model berhasil dimuat!")
-    else:
-        model = None
+    if os.path.exists(PATH_NB):
+        models["naive_bayes"] = joblib.load(PATH_NB)
+        print("Model Naive Bayes berhasil dimuat!")
+except Exception as e:
+    print(f"Error loading Naive Bayes: {e}")
+
+# Load Model LSTM
+PATH_LSTM = 'lstm_fake_news_model.h5'
+try:
+    if os.path.exists(PATH_LSTM):
+        # Gunakan joblib sesuai skrip Anda, atau load_model Keras jika error
+        models["lstm"] = joblib.load(PATH_LSTM) 
+        # models["lstm"] = load_model(PATH_LSTM)
+        print("Model LSTM berhasil dimuat!")
 except Exception as e:
-    model = None
-    print(f"Error loading model: {e}")
+    print(f"Error loading LSTM: {e}")
 
-# Skema Request dari Client
+
+# --- 2. UPDATE SKEMA REQUEST ---
 class PredictRequest(BaseModel):
     input_text: str
+    model_type: str = "naive_bayes"  # Default pakai naive_bayes jika tidak dikirim
+
 
 def scrape_berita(url):
     """Fungsi pembaca halaman web (Scraper)"""
@@ -44,26 +61,29 @@ def scrape_berita(url):
         headers = {'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'}
         response = requests.get(url, headers=headers, timeout=10)
         response.raise_for_status()
-        
         soup = BeautifulSoup(response.content, 'html.parser')
         paragraf = soup.find_all('p')
         teks_berita = " ".join([p.get_text() for p in paragraf])
-        
         return teks_berita.strip()
     except Exception as e:
         return f"GAGAL: {e}"
 
-# Endpoint Utama untuk Prediksi
+
 @app.post("/predict")
 def deteksi_hoax_api(request: PredictRequest):
-    if model is None:
-        raise HTTPException(status_code=500, detail="Model tidak ditemukan di server.")
+    # --- 3. PILIH MODEL ---
+    jenis_model = request.model_type
+    if jenis_model not in models:
+         raise HTTPException(status_code=400, detail="Pilihan model tidak valid. Gunakan 'naive_bayes' atau 'lstm'.")
+
+    aktif_model = models[jenis_model]
+    if aktif_model is None:
+        raise HTTPException(status_code=500, detail=f"Model {jenis_model} tidak ditemukan di server.")
 
     teks_mentah = request.input_text.strip()
     if not teks_mentah:
         raise HTTPException(status_code=400, detail="Input tidak boleh kosong.")
 
-    # Logika Smart Input: Cek apakah input berupa URL
     if teks_mentah.startswith("http://") or teks_mentah.startswith("https://"):
         teks_untuk_dianalisis = scrape_berita(teks_mentah)
         if teks_untuk_dianalisis.startswith("GAGAL:"):
@@ -71,21 +91,37 @@ def deteksi_hoax_api(request: PredictRequest):
     else:
         teks_untuk_dianalisis = teks_mentah
 
-    # Prediksi Probabilitas
-    proba = model.predict_proba([teks_untuk_dianalisis])[0]
-    prob_fakta = float(proba[0])
-    prob_hoax = float(proba[1])
+    # --- 4. PREDIKSI BERDASARKAN MODEL ---
+    kamus_bobot = {}
+    
+    if jenis_model == "naive_bayes":
+        # Logika untuk algoritma scikit-learn
+        proba = aktif_model.predict_proba([teks_untuk_dianalisis])[0]
+        prob_fakta = float(proba[0])
+        prob_hoax = float(proba[1])
+        
+        # Ekstraksi kata untuk highlight Frontend (biasanya hanya ada pada model linier/Naive Bayes)
+        try:
+            vec = aktif_model[0]  
+            clf = aktif_model[1]  
+            feature_names = vec.get_feature_names_out()
+            log_odds = clf.feature_log_prob_[1] - clf.feature_log_prob_[0]
+            kamus_bobot = dict(zip(feature_names, log_odds))
+        except Exception:
+            pass
+
+    elif jenis_model == "lstm":
+        # Logika untuk Deep Learning 
+        # (Perhatikan: LSTM umumnya memerlukan padding dan tokenisasi sebelum predict)
+        proba = aktif_model.predict_proba([teks_untuk_dianalisis])[0] 
+        # Jika menggunakan TensorFlow, syntaxnya mungkin berubah jadi -> aktif_model.predict(padded_sequence)[0]
+        prob_fakta = float(proba[0])
+        prob_hoax = float(proba[1])
+        # Model Deep Learning umumnya tidak memiliki "feature_log_prob", jadi highlight kata kita bypass 
+        # dan kamus_bobot dibiarkan kosong ({})
 
-    # Proses Ekstraksi Kata untuk Highlight Frontend
-    try:
-        vec = model[0]  
-        clf = model[1]  
-        feature_names = vec.get_feature_names_out()
-        log_odds = clf.feature_log_prob_[1] - clf.feature_log_prob_[0]
-        kamus_bobot = dict(zip(feature_names, log_odds))
-    except Exception:
-        kamus_bobot = {}
 
+    # Penentuan Highlight Kata (Akan berfungsi baik di Naive Bayes, dan jadi 'Netral' di LSTM jika tanpa LIME/SHAP)
     kata_kata = teks_untuk_dianalisis.split() 
     teks_highlight = []
     
@@ -93,7 +129,6 @@ def deteksi_hoax_api(request: PredictRequest):
         kata_bersih = re.sub(r'[^a-z]', '', kata.lower())
         bobot = float(kamus_bobot.get(kata_bersih, 0))
         
-        # Penentuan label per kata berdasarkan threshold
         if bobot > 0.3:
             label_kata = "Hoax"
         elif bobot < -0.3:
@@ -107,10 +142,10 @@ def deteksi_hoax_api(request: PredictRequest):
             "bobot": round(bobot, 4)
         })
 
-    # Mengembalikan response dalam format JSON
     return {
         "status": "success",
         "hasil_analisis": {
+            "model_digunakan": jenis_model,
             "teks_dianalisis": teks_untuk_dianalisis,
             "prediksi_utama": "HOAX" if prob_hoax > prob_fakta else "FAKTA",
             "probabilitas": {
@@ -119,4 +154,4 @@ def deteksi_hoax_api(request: PredictRequest):
             }
         },
         "bedah_kata": teks_highlight
-    }
+    }