Update app.py

app.py

@@ -1,48 +1,62 @@
 import gradio as gr
 import json
 import base64
+from io import BytesIO
 import requests
 import os
 
-HF_TOKEN = os.environ.get("HF_CV_ROBOT_TOKEN")  # HF Tokens
-MODEL = "Qwen/Qwen2.5-VL-7B-Instruct"  # VLM model
+# HF token & model
+HF_TOKEN = os.environ.get("HF_CV_ROBOT_TOKEN")
+MODEL = "Qwen/Qwen2.5-VL-7B-Instruct"  # 確認這個模型在 HF 支援列表裡
 
 if not HF_TOKEN:
-    print("ERROR: HF_CV_ROBOT_TOKEN is not set!")
+    print("ERROR: HF_CV_ROBOT_TOKEN environment variable not set.")
 
+# -------------------------------
+# 主處理函數 (Main Processing Function)
+# -------------------------------
 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"]
 
-        headers = {"Authorization": f"Bearer {HF_TOKEN}"}
+        # Base64 -> bytes
+        img_bytes = base64.b64decode(image_b64)
 
-        # data format
+        # 將 Base64 bytes 當作 file-like 上傳
+        files = {
+            "file": ("image.jpg", BytesIO(img_bytes), "image/jpeg")
+        }
+
+        # JSON payload 只放文字訊息
         data = {
             "model": MODEL,
             "messages": [
                 {
                     "role": "user",
                     "content": [
-                        {"type": "text", "text": "Describe this image in detail."},
-                        {"type": "image_file", "image_file": {"b64": image_b64}}
+                        {"type": "text", "text": "Describe this image in detail."}
                     ]
                 }
             ]
         }
 
-
         resp = requests.post(
             "https://router.huggingface.co/v1/chat/completions",
-            headers=headers,
-            json=data,
+            headers={"Authorization": f"Bearer {HF_TOKEN}"},
+            data={"payload": json.dumps(data)},
+            files=files,
             timeout=60
         )
 
         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}"}
 
-        # get image description as response
+        # 解析返回文字
         try:
             vlm_text = resp.json()["choices"][0]["message"]["content"][0]["text"]
         except (KeyError, IndexError, json.JSONDecodeError) as e:
@@ -55,9 +69,12 @@ 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)"),