Observation - It seems like the build agent has a skill for the OpenAI integration that lists available OpenAI models for the integration and always wants to hardcode available models in apps.
Pain Point - That creates a problem if the OpenAI integration is updated and the associated replit skill is updated, my app won’t know there’s a new model available.
User Story - I want my apps to have the ability to query available models so that I can update which model is being used in the AI configuration settings of my apps because it allows my apps to be updated to use the latest model or select from available models for a task appropriate model.
Current Workaround - I supply my own OpenAI API key for the app to query v1/models then the app will test small workloads against available models to the Replit OpenAI integration to validate whether or not those models are available then only supply available models to the application settings.