When provider_model is not specified and CHATLAS_CHAT_PROVIDER_MODEL is not set in the environment, ChatAuto always falls back to a default openai value.
I propose it would make more sense to check the environment for "known" variables and try to guess a better fallback: e.g. if OPENAI_API_KEY is defined, default to openai provider; if GOOGLE_API_KEY is defined, default to google provider, etc. If no known variable is detected, fallback to openai.
This would make it easier for an application that uses chatlas (to provide generic chat access, not tied to a particular provider) to be configured automatically based on the environment, w/out requiring an explicit configuration mechanism within the application (and without having to implement its own detection logic, that would later have to be kept in sync with chatlas' supported models).