Update app.py

app.py

@@ -10,22 +10,21 @@ MODEL = "Qwen/Qwen2.5-VL-7B-Instruct"
 def process(payload: dict):
     try:
         if not HF_TOKEN:
-            return {"error": "Missing token"}
+            return {"error": "Missing HF token"}
 
         robot_id = payload.get("robot_id", "unknown")
         image_b64 = payload["image_b64"]
 
+        # 這個是 HF 官方需要的格式
+        markdown_image = f"![](data:image/jpeg;base64,{image_b64})\nDescribe this image in detail."
+
         data = {
             "model": MODEL,
             "messages": [
                 {
                     "role": "user",
                     "content": [
-                        {"type": "text", "text": "Describe this image in detail."},
-                        {
-                            "type": "image_data",
-                            "image_data": {"b64": image_b64}
-                        }
+                        {"type": "text", "text": markdown_image}
                     ]
                 }
             ]
@@ -42,23 +41,23 @@ def process(payload: dict):
         )
 
         if resp.status_code != 200:
-            return {"error": f"VLM API error: {resp.status_code}, {resp.text}"}
+            return {"error": f"HF VLM error: {resp.status_code}, {resp.text}"}
 
         try:
-            content = resp.json()["choices"][0]["message"]["content"]
-            first = content[0]["text"]
-        except:
-            return {"error": f"Bad response: {resp.text}"}
+            vlm_text = resp.json()["choices"][0]["message"]["content"][0]["text"]
+        except Exception as e:
+            return {"error": f"Bad response: {e}, text: {resp.text}"}
 
         return {
             "received": True,
             "robot_id": robot_id,
-            "vllm_analysis": first
+            "vllm_analysis": vlm_text
         }
 
     except Exception as e:
         return {"error": str(e)}
 
+
 demo = gr.Interface(
     fn=process,
     inputs=gr.JSON(label="Input Payload"),