v4.10.0

Changelog — 4.10.0

Released: 2026-10-26

MCIdentity reads each of the nine modalities in its agents block on its own, and resolves a
global modality from MCAgents 6.3.0's per-modality settings.

Changed

• agents block (#136) — text, image, video, speech, decisions, embeddings,
  rerank, transcription and audio are each read on their own. A value other than
  global is used as written; an agents.model left over from an older config.yml is read
  as the text model.
• MCAgentsGlobal (#136) — a global modality resolves from MCAgents'
  getAgents{Modality}() getters (6.3.0), then from agents.{modality} in MCAgents' own
  config.yml, then, for the text model only, from getAgentsModel() (6.2.0) or the provider's
  default on older MCAgents.
• The plugin's agent work runs on the text model, as before (#136).
• The config.yml comment and the configuration docs no longer say the modalities are unread
  (#136).