Update app.py

app.py

@@ -1,28 +1,27 @@
 import gradio as gr
 import json
 import base64
+from io import BytesIO
 import requests
 import os
 
-# HF token & model
 HF_TOKEN = os.environ.get("HF_CV_ROBOT_TOKEN")
 MODEL = "Qwen/Qwen2.5-VL-7B-Instruct"
 
-if not HF_TOKEN:
-    print("ERROR: HF_CV_ROBOT_TOKEN environment variable not set.")
-
-# -------------------------------
-# Main Processing Function
-# -------------------------------
 def process(payload: dict):
     try:
-        if not HF_TOKEN:
-            return {"error": "Hugging Face token missing. Check environment variable."}
-
-        robot_id = payload.get("robot_id", "unknown")
         image_b64 = payload["image_b64"]
+        robot_id = payload.get("robot_id", "unknown")
+
+        # Base64 → Bytes
+        img_bytes = base64.b64decode(image_b64)
+
+        # multipart file
+        files = {
+            "file": ("image.jpg", BytesIO(img_bytes), "image/jpeg")
+        }
 
-        # 正確格式：直接把 base64 放入 image_data
+        # Router 要求的 payload 格式（不含 image_data）
         data = {
             "model": MODEL,
             "messages": [
@@ -30,7 +29,7 @@ def process(payload: dict):
                     "role": "user",
                     "content": [
                         {"type": "text", "text": "Describe this image in detail."},
-                        {"type": "image_data", "image_data": {"b64": image_b64}}
+                        {"type": "file", "file": "image.jpg"}
                     ]
                 }
             ]
@@ -39,33 +38,29 @@ def process(payload: dict):
         resp = requests.post(
             "https://router.huggingface.co/v1/chat/completions",
             headers={"Authorization": f"Bearer {HF_TOKEN}"},
-            json=data,
+            data={"payload": json.dumps(data)},
+            files=files,
             timeout=60
         )
 
         if resp.status_code != 200:
-            return {"error": f"HF VLM error: {resp.status_code}, {resp.text}"}
+            return {"error": f"VLM API error: {resp.status_code}, {resp.text}"}
 
-        try:
-            vlm_text = resp.json()["choices"][0]["message"]["content"][0]["text"]
-        except Exception as e:
-            return {"error": f"Failed to parse VLM response: {e}, Response={resp.text}"}
+        out = resp.json()
+        txt = out["choices"][0]["message"]["content"][0]["text"]
 
         return {
             "received": True,
             "robot_id": robot_id,
-            "vllm_analysis": vlm_text
+            "vllm_analysis": txt
         }
 
     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 (Dict)"),
     outputs=gr.JSON(label="Reply to Jetson"),
     api_name="predict"
 )