That is the correct library version, yes. The API for interacting with the speech recognition engine has not changed.
Numbers with a thousands separator seem to be best recognized when the command phrase also contains that separator, E.G. "10,000" rather than "10000".
Do note that depending on the locale and/or speech recognition engine, the thousands separator may be a different character (E.G. '.' instead of ','), in which case the "{TXTNUM:}" token may treat it as a decimal separator instead