Strongly agree with this request In addition to better error handling and documentation it would be very helpful to have clear rate limit and timeout behavior documented structured error codes instead of generic failures streaming examples in Python and JavaScript and best practices for production deployment This would make the model much easier to integrate into real world systems