Model
Understand model registry, capability flags, and team authorization
Models use two-level management: administrators save provider connections and model capabilities in the global model registry; then a team owner or administrator authorizes models to specific teams. Platform users only see models authorized and enabled for the current team.
Model types
Clouisle supports chat, embedding, rerank, tts, stt, audio_generation, text_to_image, and text_to_video. Different resources need matching types: Agents use chat models, Knowledge Bases use embedding and optional rerank models, and media nodes use image or video models.
Credentials and endpoints
API Keys are stored by the backend; reading a model only returns has_api_key, never plaintext credentials. Administrators can test the connection before saving; supported OpenAI-compatible endpoints are limited by the site's model endpoint whitelist.
Team quota
Authorization can configure daily/monthly Token and request limits; null means unlimited, and 0 is a valid zero limit. The higher the priority value, the earlier same-type models are sorted. Exceeding the quota returns business error 6103.
Related content: Manage models, Model provider reference.
How is this guide?