first commit

app.py

@@ -6,8 +6,7 @@ import os
 
 hf_token = os.environ["HF_TOKEN"]
 
-# Load the Gemma 3 pipeline.
-# Gemma 3 is a multimodal model that accepts text and image inputs.
+# Load the Gemma 3 pipeline
 pipe = pipeline(
     "image-text-to-text",
     model="google/gemma-3-4b-it",
@@ -15,54 +14,60 @@ pipe = pipeline(
     torch_dtype=torch.bfloat16,
     use_auth_token=hf_token
 )
+
 @spaces.GPU
 def generate_response(user_text, user_image, history):
-    messages = [
-        {
-            "role": "system",
-            "content": [{"type": "text", "text": "You are a helpful assistant."}]
-        }
-    ]
+    # Initialize message structure
+    messages = [{
+        "role": "system",
+        "content": [{"type": "text", "text": "You are a helpful assistant."}]
+    }]
+    
+    # Build user content
     user_content = []
-    if user_image is not None:
-        user_content.append({"type": "image", "image": user_image})
     if user_text:
         user_content.append({"type": "text", "text": user_text})
-    messages.append({"role": "user", "content": user_content})
+    if user_image is not None:
+        user_content.append({"type": "image", "image": user_image})
     
-    # Call the pipeline with the provided messages.
-    output = pipe(text=messages, max_new_tokens=200)
-
-    print(output)
-    print(output[0]["generated_text"][-1]["content"])
+    if user_content:
+        messages.append({"role": "user", "content": user_content})
     
-    # Attempt to extract the generated content using the expected structure.
     try:
+        # Generate response
+        output = pipe(text=messages, max_new_tokens=200)
+        
+        # Extract generated response
         response = output[0]["generated_text"][-1]["content"]
-        history.append((user_text, response))
-
-    except (KeyError, IndexError, TypeError):
-        # Fallback: return the raw output as a string.
-        #print(response)
-        pass
-        #response = str(output)
+    except Exception as e:
+        print(f"Error generating response: {e}")
+        response = "Sorry, I encountered an error processing your request."
+    
+    # Update chat history
+    if user_text or user_image:
+        history.append(((user_text, user_image) if user_image else user_text, response))
     
     return history, history
 
 with gr.Blocks() as demo:
     gr.Markdown("# Gemma 3 Chat Interface")
-    gr.Markdown(
-        "This interface lets you chat with the Gemma 3 model. "
-        "You can type a message and optionally attach an image."
-    )
-    # Specify type="messages" to avoid deprecation warnings.
-    chatbot = gr.Chatbot(type="messages")
+    gr.Markdown("This interface lets you chat with the Gemma 3 model. You can type a message and optionally attach an image.")
+    
+    chatbot = gr.Chatbot()
+    
     with gr.Row():
-        txt = gr.Textbox(show_label=False, placeholder="Type your message here...", container=False)
-        img = gr.Image(type="pil", label="Attach an image (optional)")
+        with gr.Column(scale=3):
+            txt = gr.Textbox(show_label=False, placeholder="Type your message here...", container=False)
+        with gr.Column(scale=1):
+            img = gr.Image(type="pil", label="Optional Image Upload")
+    
     state = gr.State([])
-
-    txt.submit(generate_response, inputs=[txt, img, state], outputs=[chatbot, state])
+    
+    txt.submit(
+        generate_response,
+        inputs=[txt, img, state],
+        outputs=[chatbot, state]
+    )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()