Update app.py

app.py

@@ -1,16 +1,20 @@
 import gradio as gr
 import base64
-from PIL import Image
-import io
 import json
 import requests
 
-HF_VLM_API = "https://api-inference.huggingface.co/models/Qwen/Qwen2-VL-7B-Instruct"
-HF_TOKEN = "HF_CV_ROBOT_TOKEN"  # HF Token
+HF_ROUTER_API = "https://router.huggingface.co/hf-inference"
+HF_TOKEN = "HF_CV_ROBOT_TOKEN"
+MODEL_NAME = "Qwen/Qwen2-VL-7B-Instruct"
 
 def call_vlm_api(payload: dict):
+    """
+    Call Hugging Face Router Inference API with Base64 image.
+    """
     headers = {"Authorization": f"Bearer {HF_TOKEN}"}
+    
     data = {
+        "model": MODEL_NAME,
         "inputs": [
             {
                 "image": {"b64": payload["image_b64"]},
@@ -18,16 +22,24 @@ def call_vlm_api(payload: dict):
             }
         ]
     }
-    resp = requests.post(HF_VLM_API, headers=headers, json=data, timeout=60)
-    if resp.status_code == 200:
-        return resp.json()[0].get("generated_text", "")
-    else:
-        return f"VLM API error: {resp.status_code}, {resp.text}"
+    
+    try:
+        resp = requests.post(HF_ROUTER_API, headers=headers, json=data, timeout=60)
+        if resp.status_code == 200:
+            # 取第一個 generated_text
+            return resp.json()[0].get("generated_text", "")
+        else:
+            return f"VLM API error: {resp.status_code}, {resp.text}"
+    except Exception as e:
+        return f"Exception: {str(e)}"
 
 def process(payload: dict):
+    """
+    Process JSON payload from Jetson: Base64 image + robot_id
+    Return JSON with VLM analysis
+    """
     try:
         vlm_text = call_vlm_api(payload)
-
         reply = {
             "received": True,
             "robot_id": payload.get("robot_id", "unknown"),
@@ -37,10 +49,10 @@ def process(payload: dict):
     except Exception as e:
         return {"error": str(e)}
 
-
+# Gradio MCP server
 demo = gr.Interface(
     fn=process,
-    inputs=gr.JSON(label="Input Payload (Dict format)"),
+    inputs=gr.JSON(label="Input Payload from Jetson"),
     outputs=gr.JSON(label="Reply to Jetson"),
     api_name="predict"
 )