Update app.py

app.py

@@ -67,7 +67,8 @@ with gr.Blocks() as demo:
         "<h1 style='text-align:center'>Gita Assistant (Qwen2.5-3B + LoRA)</h1>"
         "<p style='text-align:center'>Ask in English / हिंदी / ગુજરાતી. The assistant cites verses when relevant.</p>"
     )
-    system_box = gr.Textbox(
+
+    system_box  = gr.Textbox(
         value="Reply in the user’s language with 2–3 concrete points (200–400 words); cite Gita verses when relevant.",
         label="System prompt",
     )
@@ -76,11 +77,20 @@ with gr.Blocks() as demo:
     max_new     = gr.Slider(64, 1024, value=512, step=16, label="max_new_tokens")
     min_new     = gr.Slider(0, 512, value=160, step=8, label="min_new_tokens")
 
+    # IMPORTANT: keep fn=chat_fn directly, and pass components via additional_inputs
     gr.ChatInterface(
-        fn=lambda m, h: chat_fn(m, h, system_box.value, temperature.value, top_p.value, max_new.value, min_new.value),
+        fn=chat_fn,
         additional_inputs=[system_box, temperature, top_p, max_new, min_new],
-        queue=True
+        chatbot=gr.Chatbot(height=520),
+        examples=[
+            "How do I practice Nishkama Karma at work?",
+            "What does 3.19 teach about duty without attachment?",
+            "How to overcome fear of failure according to the Gita?"
+        ],
     )
 
+# For Spaces (and ZeroGPU), queue outside the interface for broad compatibility
+demo.queue(concurrency_count=2)  # keep for ZeroGPU; harmless on normal GPU
 if __name__ == "__main__":
     demo.launch()
+