capabilities.work

Download, warm, and chat with a local model on Understudy Desktop

A 8-step recipe with Understudy Labs. We can help you get access.

Recipe · 8 steps

Download, warm, and chat with a local model on Understudy Desktop

with Understudy Labs · Artificial Intelligence Management

From catalog to a locally-served conversation turn.

  1. List downloadable models and confirm the target model exists.
  2. Start downloading the chosen model.
  3. Poll the download until it completes.
  4. Add a residency slot to hold the model.
  5. Assign the downloaded model to the slot.
  6. Warm the slot so the model is resident and ready.
  7. Send the first conversation turn to the warmed model.
  8. Stream the canonical ConversationRuntime events for the run.
What you'll need to have ready
  • model_id: Catalog model to download and serve (e.g. understudy-small).
  • session_id: Conversation session identifier.
  • prompt: First user message for the warmed model.
For your developer

Arazzo workflow understudy-labs:download-warm-and-chat, 8 operations: desktopModelCatalog → desktopDownloadStart → desktopDownloadStatus → desktopResidencyAddSlot → desktopResidencyAssign → desktopResidencyWarm → conversationTurn → conversationRunEvents. Specs on APIs.io.