Upload caption_cropped_images.py with huggingface_hub

caption_cropped_images.py

@@ -9,9 +9,9 @@ import time
 import base64
 import json
 import concurrent.futures
+import requests
 from pathlib import Path
 from typing import Dict, Tuple
-from openai import OpenAI
 
 # Alibaba Qwen Configuration - Read from Environment Variables (HF Spaces Secrets)
 QWEN_API_KEY = os.environ.get("QWEN_API_KEY")
@@ -22,22 +22,10 @@ QWEN_MODEL = os.environ.get("QWEN_MODEL", "qwen-vl-max")
 if not QWEN_API_KEY:
     print("⚠️  Warning: QWEN_API_KEY not set in environment variables")
     print("   Configure in HF Spaces: Settings → Secrets")
-    qwen_client = None
+    qwen_client = "disabled"
 else:
     print(f"✓ Qwen configured: {QWEN_MODEL} at {QWEN_BASE_URL}")
-
-# Initialize client
-try:
-    if QWEN_API_KEY:
-        qwen_client = OpenAI(
-            api_key=QWEN_API_KEY,
-            base_url=QWEN_BASE_URL,
-        )
-    else:
-        qwen_client = None
-except Exception as e:
-    print(f"⚠️  Error initializing Qwen client: {e}")
-    qwen_client = None
+    qwen_client = "enabled"
 
 def encode_image(image_path: str) -> str:
     """Encode image to base64."""
@@ -55,15 +43,21 @@ def caption_single_image(image_path: str, crop_id: str) -> Tuple[str, str]:
     Returns:
         Tuple of (crop_id, caption_text)
     """
-    if not qwen_client:
+    if qwen_client != "enabled" or not QWEN_API_KEY:
         return crop_id, "unavailable"
     
     try:
         base64_image = encode_image(image_path)
         
-        response = qwen_client.chat.completions.create(
-            model=QWEN_MODEL,
-            messages=[
+        # Use requests to call the Dashscope API
+        headers = {
+            "Authorization": f"Bearer {QWEN_API_KEY}",
+            "Content-Type": "application/json",
+        }
+        
+        payload = {
+            "model": QWEN_MODEL,
+            "messages": [
                 {
                     "role": "user",
                     "content": [
@@ -78,10 +72,19 @@ def caption_single_image(image_path: str, crop_id: str) -> Tuple[str, str]:
                     ]
                 }
             ],
-            max_tokens=20
+            "max_tokens": 50
+        }
+        
+        response = requests.post(
+            f"{QWEN_BASE_URL}/chat/completions",
+            json=payload,
+            headers=headers,
+            timeout=30
         )
+        response.raise_for_status()
         
-        caption = response.choices[0].message.content.strip()
+        result = response.json()
+        caption = result["choices"][0]["message"]["content"].strip()
         return crop_id, caption
     
     except Exception as e: