max_completion_tokens forwarded unmapped — no-op until server #218 (document it)

[Khurdhula-Harshavardhan]

max_completion_tokens is forwarded unmapped — a no-op until server #218 (document it)

The SDK passes max_completion_tokens through untouched (via the ...rest spread in prepare()), but the current server reads only max_tokens. So a caller who sets only max_completion_tokens gets no output cap (default/unbounded) until server #218 deploys.

No crash — just a doc gap that over-promises. Mirror the Python SDK's resolution (interfaze-python #16): note in the README that max_completion_tokens is accepted but a no-op until server #218 ships (use max_tokens meanwhile), or add a client-side alias→max_tokens mapping (note: mapping would double-send once the server honors it natively, so the README route is preferred).

[Abhinavexist]

Decision: not adding a no-op doc note for max_completion_tokens. Server #218 is going to be merged and will honor max_completion_tokens natively — a client-side alias→max_tokens would double-send once the server reads it. So no SDK change; resolved once #218 deploys.