Update app.py

app.py

@@ -73,87 +73,31 @@ def chat_fn(message, history, system_text, temperature, top_p, max_new, min_new)
     reply = tokenizer.batch_decode(new_tokens, skip_special_tokens=True)[0].strip()
     return reply
 
-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
+# Wrap for ChatInterface + ZeroGPU
+@spaces.GPU()  # REQUIRED for ZeroGPU; remove if using standard GPU hardware
+def gradio_fn(message, history, system_text, temperature, top_p, max_new, min_new):
+    return chat_fn(message, history, system_text, temperature, top_p, max_new, min_new)
 
-    # 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=512,
-        min_new=128,
+with gr.Blocks() as demo:
+    gr.Markdown(
+        "<h1 style='text-align:center'>Gita Assistant (Qwen2.5-3B Fine-tuned)</h1>"
+        "<p style='text-align:center'>Ask in English / हिंदी / ગુજરાતી. The assistant cites verses when relevant.</p>"
     )
 
-@spaces.GPU()
-def gradio_fn(message, history):
-    response = infer_text(history + [(message, None)])
-    return response
-
-with gr.Blocks(css="""
-    .gradio-container {
-        max-width: 600px;
-        margin: auto;
-        padding: 20px;
-        font-family: sans-serif;
-        position: relative;
-        }
-    .chatbot {
-        height: 500px !important;
-        overflow-y: auto;
-        }
-    .corner {
-       position: fixed;
-       bottom: 2px;
-       z-index: 9999;
-       pointer-events: none;
-        } 
-    #left { left: 2px; }
-    #right { right: 2px; }
-    .corner img {
-       height: 500px;  /* fixed height */
-       width: auto;    /* auto to keep aspect ratio */
-        }
-    
-    """) as demo:
-    gr.Markdown(
-    """
-        <div style='text-align: center; padding: 10px;'>
-        <h1 style='font-size: 2.2em; margin-bottom: 0.2em;'>🤖 <span style='color: #4F46E5;'>kRISHNA.ai</span></h1>
-        <p style='font-size: 1.1em; color: #555;'>5000-Years of Ancient WISDOM with Modern AI ✨</p>
-        </div>
-    """,
-    elem_id="header"
+    system_box = gr.Textbox(
+        value="Reply in the user’s language with 2–3 concise points (200–400 words); cite Gita verses when relevant.",
+        label="System prompt",
     )
+    temperature = gr.Slider(0.1, 1.2, value=0.7, step=0.05, label="temperature")
+    top_p = gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="top_p")
+    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")
+
     chat = gr.ChatInterface(
         fn=gradio_fn,
-        examples=[
-            "Hello!",
-            "How can I overcome fear of failure?",
-            "How do I forgive someone who hurt me deeply?",
-            "What can I do to stop overthinking?"
-        ],
+        additional_inputs=[system_box, temperature, top_p, max_new, min_new],
         chatbot=gr.Chatbot(elem_classes="chatbot"),
-        theme="compact",
     )
-    gr.HTML(f"""
-      <div id="left" class="corner">
-        <img src="">
-      </div>
-      <div id="right" class="corner">
-        <img src="">
-      </div>
-    """)
 
 if __name__ == "__main__":
     demo.launch()