Update app.py

app.py

@@ -1,65 +1,70 @@
+import gradio as gr
+import json
 import base64
 import requests
-import json
-import io
-from PIL import Image
-import tempfile
-import time
+import os
 
-# ----------------------------
-# OpenRouter API 設定
-# ----------------------------
-OPENROUTER_KEY = "YOUR_OPENROUTER_API_KEY"
-MODEL_NAME = "qwen/qwen3-vl-32b-instruct"
-API_URL = "https://openrouter.ai/api/v1/chat/completions"
+HF_TOKEN = os.environ.get("HF_CV_ROBOT_TOKEN")
+MODEL = "Qwen/Qwen2-VL-7B-Instruct"
 
-def send_image_to_vlm(b64_image: str):
-    # 1️⃣ base64 -> PIL Image
-    img_bytes = base64.b64decode(b64_image)
-    img = Image.open(io.BytesIO(img_bytes)).convert("RGB")
+if not HF_TOKEN:
+    print("ERROR: HF_CV_ROBOT_TOKEN environment variable not set.")
 
-    # 2️⃣ 存成臨時檔
-    with tempfile.NamedTemporaryFile(suffix=".jpg") as tmp_file:
-        img.save(tmp_file.name)
+def process(payload: dict):
+    try:
+        if not HF_TOKEN:
+            return {"error": "Hugging Face token is missing. Please check Space secrets."}
 
-        # 3️⃣ 組 payload，使用 file 上傳
-        payload = {
-            "model": MODEL_NAME,
+        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}"}
+        data = {
+            "model": MODEL,
             "messages": [
                 {
                     "role": "user",
                     "content": [
-                        {"type": "text", "text": "Describe the content of this image in detail."},
-                        {"type": "file", "file": tmp_file.name}  # 這裡上傳檔案
+                        {"type": "text", "text": "Describe this image in detail."},
+                        {"type": "file", "file": image_b64}  # <- 這裡用 file 直接放 base64
                     ]
                 }
             ]
         }
 
-        headers = {
-            "Authorization": f"Bearer {OPENROUTER_KEY}",
-            "Content-Type": "application/json"
+        resp = requests.post(
+            "https://router.huggingface.co/v1/chat/completions",
+            headers=headers,
+            json=data,
+            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}"}
+
+        try:
+            vlm_text = resp.json()["choices"][0]["message"]["content"][0]["text"]
+        except (KeyError, IndexError, json.JSONDecodeError) as e:
+            return {"error": f"Failed to parse VLM response: {e}, Response text: {resp.text}"}
+
+        return {
+            "received": True,
+            "robot_id": robot_id,
+            "vllm_analysis": vlm_text
         }
 
-        # 4️⃣ 呼叫 API
-        resp = requests.post(API_URL, headers=headers, data=json.dumps(payload), timeout=120)
-        if resp.status_code == 200:
-            result = resp.json()
-            # 回傳生成的文字
-            try:
-                return result["choices"][0]["message"]["content"][0]["text"]
-            except:
-                return str(result)
-        else:
-            return f"VLM API error {resp.status_code}: {resp.text}"
+    except Exception as e:
+        print(f"An unexpected error occurred: {e}")
+        return {"error": str(e)}
 
-# ----------------------------
-# 測試
-# ----------------------------
-if __name__ == "__main__":
-    # 這裡放一張你抓到的 base64 測試
-    with open("test.jpg", "rb") as f:
-        b64_test = base64.b64encode(f.read()).decode("utf-8")
+demo = gr.Interface(
+    fn=process,
+    inputs=gr.JSON(label="Input Payload (Dict format)"),
+    outputs=gr.JSON(label="Reply to Jetson"),
+    api_name="predict"
+)
 
-    analysis = send_image_to_vlm(b64_test)
-    print("VLM 分析結果：", analysis)
+if __name__ == "__main__":
+    demo.launch(mcp_server=True)