Update app.py

app.py

@@ -4,22 +4,20 @@ import base64
 import requests
 import os
 
-HF_TOKEN = os.environ.get("HF_CV_ROBOT_TOKEN")
-MODEL = "Qwen/Qwen2-VL-7B-Instruct"
+HF_TOKEN = os.environ.get("HF_CV_ROBOT_TOKEN")  # 在 Space Secrets 裡設定
+MODEL = "Qwen/Qwen2-VL-7B-Instruct"  # 你要用的 VLM model
 
 if not HF_TOKEN:
-    print("ERROR: HF_CV_ROBOT_TOKEN environment variable not set.")
+    print("ERROR: HF_CV_ROBOT_TOKEN is not set!")
 
 def process(payload: dict):
     try:
-        if not HF_TOKEN:
-            return {"error": "Hugging Face token is missing. Please check Space secrets."}
-
         robot_id = payload.get("robot_id", "unknown")
         image_b64 = payload["image_b64"]
 
-        # Router API payload using "type": "file" for base64
         headers = {"Authorization": f"Bearer {HF_TOKEN}"}
+
+        # 這裡用 type='file' + Base64，Router API 支援直接解析
         data = {
             "model": MODEL,
             "messages": [
@@ -27,7 +25,7 @@ def process(payload: dict):
                     "role": "user",
                     "content": [
                         {"type": "text", "text": "Describe this image in detail."},
-                        {"type": "file", "file": image_b64}  # <- 這裡用 file 直接放 base64
+                        {"type": "file", "file": {"b64": image_b64, "name": "image.jpg"}}
                     ]
                 }
             ]
@@ -41,9 +39,9 @@ def process(payload: dict):
         )
 
         if resp.status_code != 200:
-            print(f"VLM API error: {resp.status_code}, {resp.text}")
             return {"error": f"VLM API error: {resp.status_code}, {resp.text}"}
 
+        # 解析回傳文字描述
         try:
             vlm_text = resp.json()["choices"][0]["message"]["content"][0]["text"]
         except (KeyError, IndexError, json.JSONDecodeError) as e:
@@ -56,9 +54,9 @@ def process(payload: dict):
         }
 
     except Exception as e:
-        print(f"An unexpected error occurred: {e}")
         return {"error": str(e)}
 
+# Gradio MCP Server
 demo = gr.Interface(
     fn=process,
     inputs=gr.JSON(label="Input Payload (Dict format)"),