¡Hello Spiritists!

This is a quick patch with only a small bugfix and a model selection extension. Here the only two points included in this patch:

- Fixed a bug that prevents candles to autoconsume on re-lighting them after being blown. Yes, candles do run out on their own after a while, some of you did notice while playing, other found out the hard way...

- Added support for GPT 4 turbo 128K, the experimental model. This model has been trained with data until April 2023, so it should result in more interesting histories and interactions.

Thanks everyone and happy boarding!