Upload agents/semantic_agent.py with huggingface_hub

agents/semantic_agent.py

@@ -18,20 +18,43 @@ def _b64(img, mx=1024):
     buf=io.BytesIO(); img.convert("RGB").save(buf,"JPEG",quality=90); return base64.b64encode(buf.getvalue()).decode()
 
 def _vlm(img, sys_prompt, user_prompt):
+    """Call VLM with deterministic timeout and single retry."""
     try:
         from openai import OpenAI
     except ImportError: return None
     token=os.environ.get("HF_TOKEN","")
     if not token: return None
-    try:
-        client=OpenAI(base_url="https://router.huggingface.co/v1",api_key=token)
-        b64=_b64(img)
-        resp=client.chat.completions.create(model="Qwen/Qwen2.5-VL-72B-Instruct",messages=[
-            {"role":"system","content":sys_prompt},
-            {"role":"user","content":[{"type":"image_url","image_url":{"url":f"data:image/jpeg;base64,{b64}"}},{"type":"text","text":user_prompt}]}
-        ],max_tokens=2000,temperature=0.1)
-        return resp.choices[0].message.content
-    except Exception as e: return f"VLM_ERROR: {e}"
+    
+    client=OpenAI(
+        base_url="https://router.huggingface.co/v1",
+        api_key=token,
+        timeout=60.0,  # Deterministic 60s timeout
+    )
+    b64=_b64(img)
+    messages=[
+        {"role":"system","content":sys_prompt},
+        {"role":"user","content":[
+            {"type":"image_url","image_url":{"url":f"data:image/jpeg;base64,{b64}"}},
+            {"type":"text","text":user_prompt}
+        ]}
+    ]
+    
+    # Try up to 2 times for deterministic behavior
+    for attempt in range(2):
+        try:
+            resp=client.chat.completions.create(
+                model="Qwen/Qwen2.5-VL-72B-Instruct",
+                messages=messages,
+                max_tokens=2000,
+                temperature=0.1,
+            )
+            return resp.choices[0].message.content
+        except Exception as e:
+            if attempt == 0:
+                import time; time.sleep(2)  # Brief pause before retry
+                continue
+            return f"VLM_ERROR: {e}"
+    return "VLM_ERROR: exhausted retries"
 
 def _parse(text):
     if not text: return {}