Speech-to-Text
Domain: Models & Providers
Motivation
@deepseek-ai/dsh-experimental-speech-to-text is the experimental service definition that selects named speech recognizers through ctx.speechToText. A defaultProvider is required and selects an exact registered id; the voice-input bundle supplies sensevoice-local. Consumers resolve their request before execution, and a missing or duplicate provider fails explicitly.
Concrete Example
defaultProvider: sensevoice-local selects the local recognizer; resolve() captures the provider instance and transcribe() rejects a withdrawn or replaced registration.
Analogy
A microphone-to-text dictation service with a pluggable recognizer engine.