Token data comes from recorded data (live or historical). What users will do almost all the time, open a model, load some recorded data, and then add new stuff to the model and configure that by also using tokens. But as those new components haven't been part of the recording, the data set doesn't show tokens. But that will be annoying to users. However, we can predict pretty nicely, that at least the tokens from the previous step will be also available for the new component. So, instead of nothing, we can at least show those predicted tokens. But we may have to provide an indicator about the fact. This screenshot is just an example, we can certainly do better than that: {width=592 height=600}