Skip to main content

The platform model marketplace and what each model costs per message

GET 

/api/ai/platform-models

Whether the workspace can run its assistant on ConnectWiz's own AI keys (key_mode: platform, billed to the wallet) instead of its own credential (byok), and which models it may pick. Each model carries an estimated wallet cost for a reference message (~900 prompt + 200 completion tokens) with the current platform markup applied, in micro-units of the wallet currency. Admin only (manage-ai-settings) and needs the ai feature (404 without it).

jurisdiction_restriction explains a model that is missing: a workspace with an active channel whose partner terms forbid a provider (today: TikTok → no DeepSeek) has that provider withdrawn on every channel, and the assistant runs on fallback_model instead.

withdrawn_selection is set when the workspace's SAVED platform model is no longer offered — its provider was withdrawn from ConnectWiz's keys (DeepSeek, since 2026-10-28) or by the rule above — and names the model the assistant runs on instead. Null when the choice stands.

byok_withdrawn is set when the workspace's OWN chat credential is of a provider its channels rule out (DeepSeek while TikTok, a Google account or a Gmail mailbox is connected). The credential is not used; the assistant runs on fallback_model on ConnectWiz's keys (billed to the wallet), or stays silent when fallback_model is null.

Responses​

The marketplace for this workspace.