Update app.py

app.py

@@ -117,52 +117,50 @@ def get_faq_from_supabase(admin_id: str) -> List[Dict]:
 
 # API Endpoints
 @app.post("/predict")
-async def predict(data: List[str] = Body(...)):
-    if len(data) != 2:
-        return {"data": ["Format input tidak valid"]}
-    
-    admin_id, question = data
-    if not question.strip():
-        return {"data": ["Pertanyaan tidak boleh kosong"]}
+async def predict(request: Request):
+    try:
+        body = await request.json()
+        admin_id = body.get("data", [None, None])[0]
+        question = body.get("data", [None, None])[1]
 
-    faqs = get_faq_from_supabase(admin_id)
-    if not faqs:
-        return {"data": ["Chatbot ini belum diisi FAQ oleh pemiliknya"]}
+        if not admin_id or not question:
+            return JSONResponse(
+                {"data": ["Admin ID atau pertanyaan tidak valid"]},
+                status_code=400
+            )
 
-    try:
-        # ... proses similarity seperti sebelumnya ...
-        if best_score < 0.3:  # Threshold lebih rendah
-            return {"data": ["Tanyakan dengan lebih spesifik tentang layanan kami"]}
-            
-        return {"data": [answers[best_idx]]}
-    except Exception:
-        return {"data": ["Sedang ada gangguan teknis, coba lagi nanti"]}
+        # Get FAQs for this admin
+        faqs = get_faq_from_supabase(admin_id)
+        if not faqs:
+            return {"data": ["Maaf, belum ada FAQ yang tersedia."]}
 
-    # Process question
-    questions = [f["question"] for f in faqs]
-    answers = [f["answer"] for f in faqs]
+        # Process question
+        questions = [f["question"] for f in faqs]
+        answers = [f["answer"] for f in faqs]
 
-    try:
-        # Encode questions and user input
-        question_embeddings = model.encode(questions, convert_to_tensor=True)
+        # Get embeddings
+        embeddings = model.encode(questions, convert_to_tensor=True)
         query_embedding = model.encode(question, convert_to_tensor=True)
         
         # Calculate similarity
-        similarity = util.pytorch_cos_sim(query_embedding, question_embeddings)
+        similarity = util.pytorch_cos_sim(query_embedding, embeddings)
         best_idx = torch.argmax(similarity).item()
         best_score = similarity[0][best_idx].item()
         
-        # Apply threshold
-        if best_score < Config.SIMILARITY_THRESHOLD:
-            return {"data": ["Maaf, saya tidak mengerti pertanyaan Anda. Coba tanyakan dengan cara lain."]}
+        # Threshold similarity (minimal 0.3)
+        if best_score < 0.3:
+            return {"data": ["Maaf, saya tidak mengerti pertanyaan Anda"]}
             
         return {"data": [answers[best_idx]]}
+
     except Exception as e:
-        raise HTTPException(
-            status_code=500,
-            detail=f"Error processing question: {str(e)}"
+        print(f"Error in prediction: {str(e)}")
+        return JSONResponse(
+            {"data": ["Terjadi kesalahan saat memproses pertanyaan"]},
+            status_code=500
         )
 
+
 @app.post("/save_chat")
 async def save_chat(chat: ChatMessage):
     """Save chat message to database"""