It can really help users out if OCS knows a few things about selected models so that we can build guardrails that will ultimately help lower frustration and make the platform for robust. See this thread as an example where it would have been useful if OCS had known what the model's token limit is.
If we know the token limit for a specific model, we can
Do do input token count and disallow users to input messages larger than that which the model can handle (on webusers though)
Do proper limiting and/or estimation for what the max token limit should be. Currently users can set this to any number, regardless of the model's context limit.
It can really help users out if OCS knows a few things about selected models so that we can build guardrails that will ultimately help lower frustration and make the platform for robust. See this thread as an example where it would have been useful if OCS had known what the model's token limit is.
Model metadata to track