Hi list, hi Abdurrahman, On Montag, 30. Oktober 2023 17:15:52 CET kevin.smith(a)isode.com wrote: Thank you for your submission, Abdurrahman. We discussed this in the Council and I want to share a few suggestions for improvement with you. The mandatory <ai> element, as Nicolas already pointed out, seems like an odd requirement: there are already LLM-based entities out there which would immediately be non-compliant. In adddition, the model selection is underspecified: How would I know which models are even available at that particular agent? I think it would be better to: - Specify a new XEP-0030 identity based on the already existing bot-identity which identifies an account as a bot. This builds on already existing mechanisms, which is always good. - Define a XEP-0030 feature or XEP-0128 service discovery extension which describes the model used by the "AI" account This allows clients/users to discover information about the specific type of model used, which is useful. - Then use separate accounts for different models. This is probably debatable; if there is a compelling usecase for hosting multiple models on the same account, the model selection could be made part of the XEP-0004 form you already suggest in the document. The available models could be made part of a XEP-0128 extension, so that clients can basically just copy the form over and select a model. Other than that, I think this is an acceptable submission. The idea of passing along parameters using an embedded form to the model is novel and a good one in my opinion. kind regards, Jonas_______________________________________________ Standards mailing list -- standards(a)xmpp.org Info: Unsubscribe: %(real_name)s-unsubscribe@%(host_name)s _______________________________________________