added keyword

app.py

@@ -54,15 +54,12 @@ def load_model_once():
     )
     logger.info(f"Model file downloaded: {model_path}")
 
-    # ---- load base model architecture ----
     logger.info("Loading base model architecture...")
     model = BertForSequenceClassification.from_pretrained(
         BASE_MODEL,
         num_labels=2,
-        # ❗ PENTING: Tidak boleh pakai low_cpu_mem_usage atau device_map
     )
 
-    # ---- load state_dict ----
     logger.info("Loading fine-tuned weights (.pth)...")
     state_dict = torch.load(model_path, map_location="cpu")
 
@@ -95,7 +92,6 @@ class StressResponse(BaseModel):
 # =====================================================
 # TWITTER API
 # =====================================================
-
 def get_user_id(username):
     url = f"https://api.x.com/2/users/by/username/{username}"
     headers = {"Authorization": f"Bearer {BEARER_TOKEN}"}
@@ -116,6 +112,26 @@ def fetch_tweets(user_id, limit=25):
     return [t["text"] for t in tweets], r.json()
 
 
+# =====================================================
+# KEYWORD EXTRACTION
+# =====================================================
+def extract_keywords(tweets):
+    stress_words = [
+        "capek", "cape", "capai", "letih", "lelah", "pusing",
+        "stress", "stres", "burnout", "kesal", "badmood",
+        "sedih", "tertekan", "muak", "bosan"
+    ]
+
+    found = set()
+    for t in tweets:
+        lower = t.lower()
+        for word in stress_words:
+            if word in lower:
+                found.add(word)
+
+    return list(found)
+
+
 # =====================================================
 # INFERENCE
 # =====================================================
@@ -162,12 +178,15 @@ def analyze(username: str):
     else:
         status = 3
 
+    keywords = extract_keywords(tweets)
+
     return StressResponse(
         message="Analysis complete",
         data={
             "username": username,
             "total_tweets": len(tweets),
             "stress_level": stress_percentage,
+            "keywords": keywords,
             "stress_status": status
         }
     )