Update backend/api.py

backend/api.py

@@ -1,12 +1,12 @@
+import pandas as pd
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
 from pydantic import BaseModel
 from transformers import AutoTokenizer
 import onnxruntime as ort
-import numpy as np # Menggantikan torch untuk operasi array
-import pandas as pd
+import numpy as np
 from pathlib import Path
-import traceback # Untuk debugging di log
+import traceback
 
 app = FastAPI()
 
@@ -22,29 +22,26 @@ app.add_middleware(
 # === Path setup ===
 BASE_DIR = Path(__file__).resolve().parent
 
-# GANTI DENGAN NAMA FILE ONNX YANG SESUAI DI FOLDER 'models'
-# Jika file Anda bernama bert_chatbot.onnx dan berada di models/
+# PERBAIKAN PATH: Sesuaikan 'bert_chatbot.onnx' dengan nama file di folder 'models'
 MODEL_PATH = BASE_DIR / "models" / "bert_chatbot.onnx"
 TOKENIZER_PATH = BASE_DIR / "models" / "bert-base-multilingual-cased" 
-DATASET_PATH = BASE_DIR / "dataset_chatbot_template.xlsx"
+DATASET_PATH = BASE_DIR / "dataset_chatbot_template.xlsx" # File ini hanya untuk melihat Intent
 
 # === Global Variables ===
 tokenizer = None
 session = None
-df_jawaban = None
+# df_jawaban tidak diperlukan untuk lookup respons, tapi kita muat untuk mencegah error.
+df_jawaban = None 
 
 # === Load tokenizer dan model ===
 try:
     print("🚀 Loading ONNX model...")
     
-    # 1. Muat Tokenizer
     tokenizer = AutoTokenizer.from_pretrained(str(TOKENIZER_PATH))
-    
-    # 2. Muat ONNX Runtime Session (Provider CPU adalah yang paling stabil di HF Free Tier)
     session = ort.InferenceSession(str(MODEL_PATH), providers=["CPUExecutionProvider"])
     
-    # 3. Muat Dataset
-    df_jawaban = pd.read_excel(DATASET_PATH)
+    # Memuat DataFrame: Hanya untuk memastikan file ada, bukan untuk lookup respons
+    df_jawaban = pd.read_excel(DATASET_PATH) 
     
     print("✅ ONNX model loaded!")
 
@@ -57,7 +54,7 @@ except Exception as e:
     pass
 
 
-# === Default responses ===
+# === Default responses (Digunakan sebagai sumber jawaban) ===
 responses = {
     "about_me": "I am a passionate developer specializing in AI and web development.",
     "skills": "My main skills are HTML5, CSS3, JavaScript, Laravel, Node.js, TensorFlow, and PyTorch.",
@@ -84,7 +81,6 @@ async def chatbot(req: ChatRequest):
         
     try:
         # 1. Tokenisasi (return_tensors="np" karena kita menggunakan NumPy/ONNX)
-        # return_tensors="np" menghemat konversi PyTorch
         inputs = tokenizer(req.text, return_tensors="np", padding=True, truncation=True, max_length=128)
         
         # 2. Dapatkan nama input dari ONNX Session
@@ -100,24 +96,22 @@ async def chatbot(req: ChatRequest):
         logits = ort_outputs[0]
         pred_id = np.argmax(logits, axis=1)[0]
         
-        # === Mapping ID ke label (ANDA HARUS TAHU MAPPING INI) ===
+        # === Mapping ID ke label ===
         id2label = {
             0: "about_me", 1: "skills", 2: "projects", 3: "experience", 
             4: "career_goal", 5: "greeting",
         }
         intent = id2label.get(pred_id, "fallback")
         
-        # === Ambil jawaban ===
-        if not df_jawaban.empty and intent in df_jawaban["Intent"].values:
-            # Menggunakan .astype(str) untuk mencegah error Pandas tipe data
-            reply = df_jawaban.loc[df_jawaban["Intent"] == intent, "Jawaban_ID"].iloc[0]
-        else:
-            reply = responses.get(intent, responses["fallback"])
-            
+        # === Ambil jawaban dari dictionary responses (Bukan DataFrame) ===
+        # Karena DataFrame Anda tidak berisi kolom Jawaban
+        reply = responses.get(intent, responses["fallback"])
+                    
         return {"reply": str(reply), "intent": intent}
         
     except Exception as e:
         import traceback
         print(f"❌ Runtime error: {e}")
         traceback.print_exc()
+        # Mengembalikan error yang valid agar frontend tahu ada masalah
         return {"reply": "⚠️ Internal server error.", "intent": "error"}