fix: bound Work model generation

agent_platform/runtime/provider.py

@@ -9,6 +9,7 @@ from agent_platform.config import Settings
 
 TRANSIENT_COMPLETE_STATUS_CODES = frozenset({408, 425, 429, 500, 502, 503, 504, 524})
 COMPLETE_MAX_ATTEMPTS = 2
+COMPLETE_MAX_TOKENS = 4096
 MAX_RETRY_DELAY_SECONDS = 5.0
 
 
@@ -152,6 +153,7 @@ class ModelProvider:
             "model": model,
             "messages": messages,
             "stream": True,
+            "max_tokens": COMPLETE_MAX_TOKENS,
         }
         if tools:
             payload["tools"] = tools