fix: use MODEL_LOG env var instead of hardcoded onstart.log to prevent log scanner feedback loop

workers/openai/worker.py

@@ -10,7 +10,7 @@ logging.getLogger().setLevel(logging.WARNING)  # Only show warnings and errors
 # Ollama model configuration
 MODEL_SERVER_URL           = 'http://127.0.0.1'
 MODEL_SERVER_PORT          = 11434
-MODEL_LOG_FILE             = '/var/log/onstart.log'
+MODEL_LOG_FILE             = os.environ.get('MODEL_LOG', '/var/log/ollama.log')
 MODEL_HEALTHCHECK_ENDPOINT = "/"
 
 # Ollama-specific log messages