Update app.py

app.py

@@ -10,11 +10,11 @@ APIFY_TOKEN = os.getenv("APIFY_TOKEN")
 HF_TOKEN = os.getenv("HF_TOKEN")
 GROQ_API_KEY = os.getenv("GROQ_API_KEY")
 
-# Validasi token saat startup
+# Validasi token saat startup (Gemini sudah dihapus)
 if not HF_TOKEN:
     print("⚠️ WARNING: HF_TOKEN is not set! Sentiment analysis will fail.")
-if not GEMINI_KEY:
-    print("⚠️ WARNING: GEMINI_KEY is not set! Gemini insights will fail.")
+if not GROQ_API_KEY:
+    print("⚠️ WARNING: GROQ_API_KEY is not set! Llama 3 insights will fail.")
 if not APIFY_TOKEN:
     print("⚠️ WARNING: APIFY_TOKEN is not set! Scraping will fail.")
 
@@ -31,7 +31,7 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# Hugging Face — nlptown model (always active, 1-5 star output, supports Indonesian)
+# Hugging Face — nlptown model
 HF_API_URL = "https://router.huggingface.co/hf-inference/models/nlptown/bert-base-multilingual-uncased-sentiment"
 HF_HEADERS = {"Authorization": f"Bearer {HF_TOKEN}"}
 
@@ -68,13 +68,11 @@ def analyze_sentiment(ulasan_list):
                 
                 # nlptown returns [[{"label": "5 stars", "score": 0.5}, ...]]
                 if isinstance(res, list) and len(res) > 0:
-                    # Get the top prediction
                     predictions = res[0] if isinstance(res[0], list) else res
                     if predictions and isinstance(predictions[0], dict):
                         top_label = predictions[0].get("label", "3 stars")
                         stars = int(top_label[0])  # "5 stars" → 5
                         
-                        # Map stars to sentiment: 1-2=neg, 3=neu, 4-5=pos
                         if stars >= 4:
                             sentiment = "pos"
                         elif stars <= 2:
@@ -93,8 +91,7 @@ def analyze_sentiment(ulasan_list):
                         
                         for asp in matched_aspects:
                             raw_scores[asp][sentiment] += 1
-                        
-                        print(f"✅ Sentiment: {stars} stars → {sentiment} | Aspects: {[a[:4] for a in matched_aspects]}")
+                            
                     else:
                         print(f"HF unexpected format: {str(predictions)[:100]}")
                 else:
@@ -110,7 +107,6 @@ def analyze_sentiment(ulasan_list):
 def get_advanced_insights(ulasan_list):
     try:
         gabungan_ulasan = " | ".join(ulasan_list[:8])
-        # Prompt disesuaikan sedikit agar Llama 3 lebih menurut
         prompt = f"""
         Kamu adalah analis bisnis restoran. Analisis ulasan berikut: {gabungan_ulasan}.
         Kamu WAJIB mengembalikan respon HANYA dalam format JSON murni ini:
@@ -129,9 +125,9 @@ def get_advanced_insights(ulasan_list):
         }
         
         payload = {
-            "model": "llama3-8b-8192", # Model Llama 3 yang sangat cepat
+            "model": "llama3-8b-8192", 
             "messages": [{"role": "user", "content": prompt}],
-            "response_format": {"type": "json_object"}, # Memaksa output berupa JSON
+            "response_format": {"type": "json_object"}, 
             "temperature": 0.5
         }
         
@@ -139,7 +135,6 @@ def get_advanced_insights(ulasan_list):
         
         if response.status_code == 200:
             data = response.json()
-            # Parsing output khas arsitektur OpenAI/Groq
             t = data["choices"][0]["message"]["content"]
             return json.loads(t)
         else:
@@ -152,7 +147,7 @@ def get_advanced_insights(ulasan_list):
             "kata_kunci_negatif": ["Server Sibuk"], 
             "rekomendasi": [{"judul": "Gunakan Fallback", "deskripsi": "Server LLM sedang sibuk.", "prioritas": "Low"}]
         }
-        
+
 # --- 4. ENDPOINT ---
 @app.post("/analyze")
 async def analyze_realtime(request: Request):
@@ -160,7 +155,7 @@ async def analyze_realtime(request: Request):
         body = await request.json()
         target_url = body.get("url")
         
-        # Scraper (15 ulasan — sesuaikan dengan batas proses AI)
+        # Scraper
         run = apify_client.actor("compass/google-maps-reviews-scraper").call(
             run_input={"startUrls": [{"url": target_url}], "maxReviews": 100, "language": "id"}
         )
@@ -181,7 +176,7 @@ async def analyze_realtime(request: Request):
 
         # Proses AI
         raw_scores = analyze_sentiment(ulasan_teks_saja)
-        gemini_data = get_advanced_insights(ulasan_teks_saja)
+        insights_data = get_advanced_insights(ulasan_teks_saja) # Variabel diganti dari gemini_data
         
         # Kalkulasi
         sentiment_results = []
@@ -208,11 +203,11 @@ async def analyze_realtime(request: Request):
                 "overall_sentiment": {"positive": over_p, "neutral": over_neu, "negative": over_n},
                 "sentiment_score": sentiment_results,
                 "kata_kunci": {
-                    "positif": gemini_data.get("kata_kunci_positif", []),
-                    "negatif": gemini_data.get("kata_kunci_negatif", [])
+                    "positif": insights_data.get("kata_kunci_positif", []),
+                    "negatif": insights_data.get("kata_kunci_negatif", [])
                 },
                 "contoh_review": ulasan_lengkap[:6],
-                "rekomendasi_list": gemini_data.get("rekomendasi", [])
+                "rekomendasi_list": insights_data.get("rekomendasi", [])
             }
         }
     except Exception as e: