Update src/qa.py

src/qa.py

@@ -160,20 +160,29 @@ def generate_answer(query: str, retrieved_chunks: list, reasoning_mode: bool = F
     )
 
     try:
+        # 🧠 Adaptive length for factual mode (reasoning_mode=False)
+        if reasoning_mode:
+            max_tokens = 180  # keep reasoning slightly longer
+        else:
+            max_tokens = 120 if len(query.split()) < 6 else 180  # short Qs stay fast
+
         result = _answer_model(
             prompt,
-            max_new_tokens=180 if reasoning_mode else 120,
+            max_new_tokens=max_tokens,
             temperature=0.6 if reasoning_mode else 0.3,
             do_sample=reasoning_mode,
             early_stopping=True,
             pad_token_id=_tokenizer.eos_token_id,
         )
+
         text = result[0]["generated_text"].strip()
         return text.split("Answer:")[-1].strip() if "Answer:" in text else text
+
     except Exception as e:
         print(f"⚠️ Generation failed: {e}")
         return "⚠️ Error: Could not generate an answer."
 
+
 # ==========================================================
 # 7️⃣ Local Test
 # ==========================================================