dsh-voice-mic: Add voice input for DSH Web GUI
`dsh-voice-mic` is a voice input plugin for DSH Web GUI, designed to streamline the input process for long text or code. It provides a microphone button and an `Alt+V` shortcut key, supporting real-time transcription of recordings written to the input box draft. It does not automatically send messages, ensuring users confirm the content. The plugin supports three recognition engines: first, browser-based recognition using the Web Speech API, suitable for quick testing; second, a local offline backend supporting SenseVoiceSmall and Paraformer models, bound to `127.0.0.1:7860` by default to ensure audio does not leave the local machine, requiring a Python 3.9+ environment; and third, a cloud API compatible with the OpenAI interface, where real-time transcription incurs corresponding fees. Installation requires DSH 0.1.0-
Read Moredsh-voice: The Voice Input Plugin for DeepSeek Harness
dsh-voice is a voice input plugin for DeepSeek Harness, maintained by 3274375092 under the MIT license. Its npm package name is @nn12138/dsh-voice. It allows users to dictate directly via microphone in web chat boxes, submitting recognized text as regular chat messages without altering the agent preset or system prompt throughout the process. The plugin supports real-time streaming transcription, prioritizing the host's native ASR (sherpa-onnx-node) and automatically falling back to browser Web Speech when unavailable. Core features include a microphone button, configurable global hotkey (default Ctrl+Space), and adaptive dual-engine recognition. Installation requires Node ^22.19.0 or higher.
Read More