Update app.py

app.py

@@ -1,12 +1,11 @@
 import os
-import zipfile
+import pickle
 import numpy as np
 import tensorflow as tf
 from fastapi import FastAPI
 from pydantic import BaseModel
 from fastapi.middleware.cors import CORSMiddleware
 
-# Optimasi RAM server HF
 os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
 os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
 
@@ -19,80 +18,43 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-clf_model = None
-
-# Schema request dari Space A
 class ItemRequest(BaseModel):
     items_text: str
 
-# =========================================================
-# JURUS PAMUNGKAS: MANUAL BYPASS + ZIP EXTRACT
-# =========================================================
-def build_manual_clf_model():
-    inputs = tf.keras.Input(shape=(1,), dtype=tf.string, name='input_layer')
-    
-    x = tf.keras.layers.TextVectorization(
-        max_tokens=5000, 
-        output_sequence_length=50, 
-        name='text_vectorization'
-    )(inputs)
-    
-    x = tf.keras.layers.Embedding(input_dim=5000, output_dim=64, name='embedding')(x)
-    x = tf.keras.layers.GlobalAveragePooling1D(name='global_average_pooling1d')(x)
-    x = tf.keras.layers.Dense(64, activation='relu', name='dense')(x)
-    x = tf.keras.layers.Dropout(0.2, name='dropout')(x)
-    
-    outputs = tf.keras.layers.Dense(7, activation='softmax', name='dense_1')(x)
-    
-    return tf.keras.Model(inputs=inputs, outputs=outputs, name='functional')
+model = None
+encoder = None
 
-def get_clf_model():
-    global clf_model
-    if clf_model is None:
-        print("Membangun arsitektur manual dan memuat bobot...")
-        # 1. Bangun arsitektur manual untuk menghindari quantization_config
-        clf_model = build_manual_clf_model()
-        # 2. Hanya muat bobot angka
-        clf_model.load_weights("model_strukly.keras")
-        
-        # 3. Suntikkan kamus TextVectorization langsung dari dalam file .keras
-        print("Mengekstrak kamus TextVectorization...")
-        tv_layer = clf_model.get_layer('text_vectorization')
-        try:
-            with zipfile.ZipFile("model_strukly.keras", 'r') as z:
-                # Cari file txt di folder assets
-                vocab_files = [n for n in z.namelist() if 'assets/' in n and n.endswith('.txt')]
-                if vocab_files:
-                    with z.open(vocab_files[0]) as f:
-                        vocab_list = f.read().decode('utf-8').splitlines()
-                        tv_layer.set_vocabulary(vocab_list)
-                        print(f"Kamus berhasil disuntikkan dari: {vocab_files[0]}")
-        except Exception as e:
-            print(f"Gagal menyuntikkan kamus: {e}")
-            
-    return clf_model
+def get_model():
+    global model, encoder
+    if model is None:
+        model = tf.keras.models.load_model('model_strukly.keras')
+        with open('penerjemah_kategori.pkl', 'rb') as file:
+            encoder = pickle.load(file)
+    return model, encoder
 
 @app.post("/classify/")
 def classify_items(request: ItemRequest):
     try:
         raw_items = request.items_text
         
-        # Jika Space A mengirim string kosong
         if not raw_items.strip():
-             return {"success": True, "class_id": None}
+             return {"success": True, "category_id": None, "category_name": "Tidak Terdeteksi"}
              
-        model = get_clf_model()
-        input_data = np.array([raw_items])
+        m, e = get_model()
         
-        # Prediksi dengan model AI
-        hasil_prediksi = model(input_data, training=False).numpy()
+        input_data = tf.constant([raw_items], dtype=tf.string)
+        hasil_prediksi = m.predict(input_data, verbose=0)
         
-        # Ambil indeks dengan probabilitas tertinggi
-        class_id = int(np.argmax(hasil_prediksi[0]))
+        category_id = int(np.argmax(hasil_prediksi[0]))
+        category_name = e.inverse_transform([category_id])[0]
         
-        return {"success": True, "class_id": class_id}
+        return {
+            "success": True, 
+            "class_id": category_id,
+            "class_name": category_name
+        }
         
     except Exception as e:
         import traceback
-        print(f"Error AI Klasifikasi detail: {traceback.format_exc()}")
-        return {"success": False, "error": str(e), "class_id": None}
+        print(f"Error AI Klasifikasi: {traceback.format_exc()}")
+        return {"success": False, "error": str(e), "category_id": None, "category_name": None}