dsh-speech: DeepSeek Harness Voice Service Plugin

dsh-speech is a speech processing plugin for DeepSeek Harness (DSH). It provides Token-gated speech recognition (ASR) and synthesis (TTS) services through standalone routing without modifying the host source code. The plugin supports multiple cloud providers, such as Alibaba Cloud Bailian, OpenAI, Groq, and SiliconFlow, as well as local services such as SenseVoice and CosyVoice2, and features capabilities including streaming transcription and real-time voice activity detection. Users can flexibly combine ASR and TTS providers to achieve private deployment or multi-cloud switching, making it suitable for scenarios such as offline speech processing.

Read More