Skip to main content

Crate onde

Crate onde 

Source
Expand description

onde — On-device inference abstraction layer for SplitFire AI.

This crate centralises all mistral.rs-backed model management so that individual Tauri apps remain thin command wrappers:

  • inference::models — model ID constants and rich metadata used across the app (download size, display name, org, description).
  • inference::token — HuggingFace token resolution (build-time literal or on-disk cache; required on iOS where the filesystem is sandboxed).
  • hf_cache — HuggingFace hub cache inspection, repair, and model download with a progress-callback API that is decoupled from Tauri.

§Re-exports

mistralrs, hf_hub, and mistralrs_core are re-exported so that apps depending on onde do not need their own direct dependency on those crates. Access them as onde::mistralrs, onde::hf_hub, and onde::mistralrs_core.

Modules§

hf_cache
HuggingFace hub cache inspection, repair, and model download.
inference