dsh-tts-service: Connecting DSH Agents to Multi-vendor Voice Synthesis

dsh-tts-service is a multi-vendor TTS plugin for DeepSeek Harness, maintained by loveandbeepbrother. The package provides six independently switchable plugin entry points: tts-service provides the ctx.ttsService service (vendor registration, voice selection, synthesis, and voice list), tts-doubao connects to Volcengine Doubao TTS 2.0, tts-bailian connects to Alibaba Cloud Bailian (CosyVoice, Qwen-Audio-TTS, Qwen3-TTS), tts-openai-compatible interfaces with any endpoint compatible with POST /audio/speech, tts-say exposes agent tools such as tts_service_say, tts_service_voices, and tts_service_select, and tts-speaker-windows plays audio through the system speaker on Windows. The plugin has no default vendor or default voice; if not selected, tts_service_say returns VOICE_NOT_SELECTED. The current version only synthesizes complete audio segments and does not support streaming output. Installation can be performed using `dsh plugin add @loveandbeepbrother/dsh-tts-service` or the corresponding `github:` format, requiring DSH >=0.2.1-rc.3 <0.3.0-0 and Node.js 22+. Key names are stored, but values are not; values are retrieved from the DSH credential service or startup environment variables and read on every request. Voice selection is stored in the tts-service settings and can be configured via the settings page, agent tools, or ctx.ttsService.configure, taking effect immediately; error handling is explicit to prevent silent vendor switching. The article also lists default configuration fields and tone description adherence for two vendors, as well as notes on licensing, version numbers, and precautions such as "run with process permissions and check source code before installation," and provides links to the community directory page and GitHub.

Read More