Upload agents/semantic_agent.py with huggingface_hub

agents/semantic_agent.py

@@ -53,8 +53,12 @@ def _vlm(img, sys_prompt, user_prompt):
         except Exception as e:
             last_error = e
             if attempt < 2:
+                err_str = str(last_error)
+                # Don't retry on payment/quota errors — it won't help
+                if '402' in err_str or 'credit' in err_str.lower() or 'quota' in err_str.lower():
+                    return f"VLM_ERROR: Inference credits depleted. Add HF Pro subscription or purchase credits at huggingface.co/settings/billing"
                 import time
-                wait = 3 * (attempt + 1)  # 3s, 6s backoff
+                wait = 3 * (attempt + 1)
                 time.sleep(wait)
                 continue
     return f"VLM_ERROR: {last_error}"