Update app.py

app.py

@@ -9,9 +9,79 @@ import asyncio
 from dotenv import load_dotenv
 import ast # Import the Abstract Syntax Tree module
 
-# ... (imports and global variables remain the same) ...
+# Load environment variables (ensure .env is set up locally)
+load_dotenv()
 
-# ... (process_webcam_stream_async function remains the same) ...
+ROBOT_ID = "Robot_MCP_Client"
+HF_TOKEN = os.environ.get("HF_TOKEN")
+if not HF_TOKEN:
+    print("Warning: HF_TOKEN not found. API calls may fail.")
+    HF_TOKEN = "missing_token_placeholder" 
+
+# The MCP URL of your remote server
+MCP_SERVER_URL = "https://oppaai-robot-mcp-server.hf.space/gradio_api/mcp/"
+SERVER_NAME = "Robot_MCP_Server"
+TOOL_NAME = "Robot_MCP_Server_robot_watch" 
+
+
+# Initialize the MCP client globally
+HTTP_TRANSPORT = StreamableHttpTransport(url=MCP_SERVER_URL)
+MCP_CLIENT = Client(transport=HTTP_TRANSPORT, name=SERVER_NAME)
+
+
+async def process_webcam_stream_async(image):
+    """Send webcam image to HF MCP Server using MCP protocol and get result"""
+    if image is None:
+        return "", "", "", ""
+
+    if HF_TOKEN == "missing_token_placeholder":
+        return "Error: HF_TOKEN not set locally.", "", "", ""
+
+    buffered = io.BytesIO()
+    image.save(buffered, format="JPEG")
+    b64_img = base64.b64encode(buffered.getvalue()).decode("utf-8")
+
+    payload = {
+        "hf_token_input": HF_TOKEN,
+        "robot_id_input": ROBOT_ID,
+        "image_b64_input": b64_img
+    }
+
+    try:
+        async with MCP_CLIENT:
+            response = await MCP_CLIENT.call_tool(TOOL_NAME, payload)
+            
+            if response.is_error:
+                error_text = response.content[0].text if response.content and isinstance(response.content, list) else "Unknown error"
+                raise Exception(f"MCP Tool Error: {error_text}")
+
+            # Use ast.literal_eval because your server returns Python-formatted strings (single quotes)
+            raw_text = response.content[0].text
+            response_dict = ast.literal_eval(raw_text)
+            
+            # --- 👇 EXTRACTING EACH FIELD CORRECTLY 👇 ---
+            vlm_result = response_dict.get("result", {})
+            
+            description_out = vlm_result.get("description", "")
+            human_out = vlm_result.get("human", "")
+            objects_list = vlm_result.get("objects", [])
+            environment_out = vlm_result.get("environment", "")
+
+            # Format the objects list into a single string for display
+            objects_str = ", ".join(objects_list) if isinstance(objects_list, list) else str(objects_list)
+
+            return (
+                description_out,
+                human_out,
+                objects_str,
+                environment_out
+            )
+            
+    except Exception as e:
+        print(f"Error calling remote MCP API: {e}") 
+        import traceback
+        traceback.print_exc() # Helps debug other errors
+        return f"Error: {e}", "", "", ""
 
 with gr.Blocks() as demo:
     gr.Markdown("## 🎥 Robot Vision Webcam Stream (using MCP Client)")