Voice notes in, spoken answers out — dictate audio that becomes user messages (transcribe), have the agent read replies aloud (speak), and leave walk-away narration on long headless runs. Local-first: plain audio files under ~/.dsh/voice/.
Resilient multi-provider web search & content extraction for Claude Code, DeepSeek Harness, and any MCP host — quota-aware failover, JS rendering, and a CLI.
Generic stateless dialogue compression that mimics human memory. LLM conversations stay bounded forever — no session management, no context overflow. Ships a ready-to-use adapter for DeepSeek Harness (DSH).
DeepSeek Harness (dsh web) 语音播报插件:用 Edge TTS 微软神经语音朗读每个助手回复的最终结果。只读最终答复、长文自动分段、输入框喇叭一键开关、审批提示朗读与自定义提示音,全程本地零第三方依赖。 Text-to-speech for dsh web via Edge TTS.
Universal message-platform gateway for the DSH web GUI: WeCom AI bot / Telegram / Discord / QQ / Email bridges with per-chat sessions, streaming replies, and auto context compression
Speech-to-text voice input plugin for DeepSeek Harness (DSH) web GUI: click the mic in the composer to turn speech into text in the input box. Browser Web Speech API + OpenAI-compatible Whisper (OpenAI/Groq) with selectable model. DSH 语音输入插件
Give DeepSeek (text-only) models **vision** in Claude Code and Codex by routing image files to any OpenAI-compatible vision endpoint (OpenRouter, SiliconFlow, DashScope, Ollama, llama.cpp, vLLM, LM Studio, …). Zero runtime dependencies, MIT licensed.
DeepSeek Harness plugin for local image / voice / music / SFX generation: the same character stays the same character across calls, degenerate output is refused rather than returned, and the GPU is untouched when idle. Hand the image half to any OpenAI-shaped API instead and its engine and 10.1 GiB of weights are never installed.