Update app.py

app.py

@@ -63,6 +63,27 @@ def chat_fn(message, history, system_text, temperature, top_p, max_new, min_new)
     if eos:
         gen_cfg_kwargs["eos_token_id"] = eos
 
+    def infer_text(history, system_text=""):
+    """
+    Reply in the user’s language with 2–3 concise points (200–400 words); cite Gita verses when relevant.
+    """
+    if not history:
+        return ""  # nothing to answer
+
+    # Split out the newest user message and the prior turns
+    last_user_msg, _ = history[-1]
+    prior_history = history[:-1]
+
+    # Call your existing generator with sane defaults
+    return chat_fn(
+        message=last_user_msg,
+        history=prior_history,
+        system_text=system_text,
+        temperature=0.7,
+        top_p=0.9,
+        max_new=256,
+    )
+
     gen_cfg = GenerationConfig(**gen_cfg_kwargs)
 
     with torch.no_grad():