Mirai uzu Inference SDK

uzu is Mirai's first-party, high-performance on-device inference engine. A single high-level API — shared across Python, Swift, TypeScript, and Rust — downloads and runs LLM, text-to-speech, and classification models locally, with speculative decoding, structured output, and an optional cloud path.

API entry from apis.yml

apis.yml Raw ↑
name: Mirai uzu Inference SDK
description: uzu is Mirai's first-party, high-performance on-device inference engine. A single high-level
  API — shared across Python, Swift, TypeScript, and Rust — downloads and runs LLM, text-to-speech, and
  classification models locally, with speculative decoding, structured output, and an optional cloud path.
humanURL: https://docs.trymirai.com
tags:
- On-Device AI
- Inference
- SDK
properties:
- type: Documentation
  url: https://docs.trymirai.com
- type: GettingStarted
  url: https://docs.trymirai.com/quick-start/chat
- type: Packages
  url: packages/mirai-packages.yml
- type: SDKs
  url: packages/mirai-packages.yml
- type: LLMsTxt
  url: llms/mirai-llms.txt
- type: SourceCode
  url: https://github.com/trymirai/uzu