DeepSeek Harness plugins — Page 967

Lightweight realtime speech-to-speech for DeepSeek Harness with Qwen and OpenAI providers.

Realtime Avatar (realtimeavatar.ai) for the DeepSeek Harness: the developer's API key held by the harness, the public docs as on-demand skills, rta_* tools over the public REST API, and a /rta command that walks a developer from no key to a live avatar call.

Orphan process and port reaper for DeepSeek Harness: find the dev servers, watchers and background processes an agent left behind, attribute them to a workspace, preview the blast radius, and reclaim the ports.

DeepSeek Harness per-model thinking-depth (reasoning effort) configurator: a settings-page section that auto-reads the configured models from llm-pi-ai and applies a per-model thinking level through the agent/request waterfall — all without modifying DSH itself.

Distill over-long model reasoning into its useful core and replay the distilled version on every later request, with a per-session cost report.

Add configurable thinking-strength (reasoning effort) levels to custom llm-pi-ai providers in DeepSeek Harness

Codex-style DeepSeek Harness model and reasoning selector with model-advertised effort levels, DSH-native themes, left-clipped radiation effects, and copy-ready reasoning-effort declaration guidance for custom-provider models.

Per-model reasoning-effort editor with embedded slider for DeepSeek Harness custom providers.

A draggable off / low / high / max reasoning-effort slider in the DeepSeek Harness composer model row, bound to the real per-session model selection.

Frontend-injected per-model reasoning-effort (thinking strength) configuration for custom pi-ai providers. Registers a per-model 思考强度 editor into the Settings -> Models provider card and persists reasoningEfforts through the settings Remote.

Codex-style reasoning-effort slider for DeepSeek Harness: turns the model menu's Effort row into a continuous, level-snapping slider.

给第三方 provider 的模型按关键词自动挂上思考强度档位（off/low/medium/high/max），新增模型后无需手动配置

Codex-style folding for the DeepSeek Harness web transcript: a running turn keeps only its last few actions visible, folded behind one row that opens the whole turn.

Make the DSH agent's thinking process (reasoning) use a language you choose — Simplified Chinese by default — and expand thinking rows by default. / 让 DSH 的思考过程用指定语言（默认简体中文）输出，并把思考行默认展开。

Detect and abort repeating reasoning (thinking-loop stalls) in DSH, plus a fire journal and maintenance tool.

DeepSeek Harness 推理强度与网关增强插件：自动为 llm-pi-ai 下所有模型补齐推理强度档位（off→max 全七档），并自动为 OpenCode Go 网关注入必需的 x-opencode-session 路由 Header，写入 settings 由 dsh 原生解析生效。

推理等级++：为 API 地址非官方的自定义提供方补齐与官方同形的推理等级，且每一档在上游真实生效（经实测探测，不可区分的档位不出现）

A minimal reasoning-effort ruler for the DSH composer: one hairline, a sliding marker, per-model memory, optimistic switching — and a streamlined model picker.

Reasoning-effort setup page for DeepSeek Harness: detects models that expose no thinking levels and edits their llm-pi-ai reasoningEfforts declaration from a settings page

DSH Web plugin: replace the composer reasoning-effort (推理等级) picker with an inline slider card.

四档推理强度滑块与模型选择器 · DeepSeek Harness UI 插件

DeepSeek Harness 插件：把「推理等级」(reasoning effort) 档位名翻译成中文，并为接入的自定义第三方模型注入推理等级显示（官方 DSH 默认不注入），所选档位写入出站请求体。

Reaudit MCP for DeepSeek Harness: mounts the hosted Reaudit MCP server (https://mcp.reaudit.io/mcp) through the in-box @deepseek-ai/dsh-mcp-client bridge.

DeepSeek Harness 跨会话记忆（按需）：列出并加载之前会话干了什么，不自动污染上下文。Cross-session memory for DSH, loaded on demand.