Update app.py

app.py

@@ -1,45 +1,20 @@
 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")
 MODEL = "Qwen/Qwen2.5-VL-7B-Instruct"
 
-HF_UPLOAD_URL = "https://huggingface.co/api/uploads"
-
-def upload_to_hf(bytes_data):
-    """Upload image bytes to HF and return image_url."""
-    resp = requests.post(
-        HF_UPLOAD_URL,
-        headers={"Authorization": f"Bearer {HF_TOKEN}"},
-        files={"file": ("temp.jpg", bytes_data, "image/jpeg")}
-    )
-
-    if resp.status_code != 200:
-        raise RuntimeError(f"HF upload failed: {resp.text}")
-
-    url = resp.json()["url"]
-    return url
-
-
 def process(payload: dict):
     try:
         if not HF_TOKEN:
-            return {"error": "Missing HF token."}
+            return {"error": "Missing token"}
 
         robot_id = payload.get("robot_id", "unknown")
-
-        # --- get image bytes
         image_b64 = payload["image_b64"]
-        img_bytes = base64.b64decode(image_b64)
 
-        # --- upload to HF (get public URL)
-        image_url = upload_to_hf(img_bytes)
-
-        # --- VLM request (image_url only)
         data = {
             "model": MODEL,
             "messages": [
@@ -47,7 +22,10 @@ def process(payload: dict):
                     "role": "user",
                     "content": [
                         {"type": "text", "text": "Describe this image in detail."},
-                        {"type": "image_url", "image_url": {"url": image_url}}
+                        {
+                            "type": "image_data",
+                            "image_data": {"b64": image_b64}
+                        }
                     ]
                 }
             ]
@@ -55,7 +33,10 @@ def process(payload: dict):
 
         resp = requests.post(
             "https://router.huggingface.co/v1/chat/completions",
-            headers={"Authorization": f"Bearer {HF_TOKEN}", "Content-Type": "application/json"},
+            headers={
+                "Authorization": f"Bearer {HF_TOKEN}",
+                "Content-Type": "application/json"
+            },
             data=json.dumps(data),
             timeout=60
         )
@@ -64,24 +45,24 @@ def process(payload: dict):
             return {"error": f"VLM API error: {resp.status_code}, {resp.text}"}
 
         try:
-            vlm_text = resp.json()["choices"][0]["message"]["content"][0]["text"]
+            content = resp.json()["choices"][0]["message"]["content"]
+            first = content[0]["text"]
         except:
-            return {"error": f"Bad VLM response: {resp.text}"}
+            return {"error": f"Bad response: {resp.text}"}
 
         return {
             "received": True,
             "robot_id": robot_id,
-            "vllm_analysis": vlm_text
+            "vllm_analysis": first
         }
 
     except Exception as e:
         return {"error": str(e)}
 
-
 demo = gr.Interface(
     fn=process,
-    inputs=gr.JSON(label="Input Payload (Dict format)"),
-    outputs=gr.JSON(label="Reply to Jetson"),
+    inputs=gr.JSON(label="Input Payload"),
+    outputs=gr.JSON(label="Reply"),
     api_name="predict"
 )