dsh-voice

When using DeepSeek Harness for conversations, manual typing can interrupt your flow or reduce efficiency, especially on the go or when typing is inconvenient. The dsh-voice plugin addresses this by enabling voice input that directly converts your speech into chat messages, letting you focus on your thoughts rather than keystrokes. Its core strength lies in the adaptive dual-engine system: it uses offline native ASR (sherpa-onnx-node) when a local model is available for privacy and speed, but gracefully falls back to the browser Web Speech API for zero-config operation otherwise. It offers real-time partial transcripts, hotkey activation, and remains preset-agnostic, acting merely as an alternative input method. This plugin is ideal for all DeepSeek Harness users—developers, content creators, or casual chatters—who seek a more natural and efficient way to interact without compromising on features or privacy.