Plugins

Persistent memory for DeepSeek Harness: memory read/write/retrieve tools, per-turn auto-recording, project-scoped recall injection and a management UI, with non-destructive Claude Code memory import.

Inline image previews in chat for pasted images bridged by dsh-tool-vision on text-only models (click to zoom, old messages get previews too).

Condenses DSH session titles from a full sentence into a short verb-free phrase, aiming for under five words.

Usage statistics panel: silent per-model token collection, activity heatmap, trend charts, cost estimation and daily-limit guard.

Extends the DSH web chat drag-and-drop to accept every file type: non-image files are saved into the active session workspace so the agent can read and use them.

Enables or disables MCP servers from Settings → MCP Servers: stops or starts each @deepseek-ai/dsh-mcp-client connection immediately and persists the change across restarts.

Enables or disables DSH plugins from Settings → Plugins: each toggle stops or starts the plugin immediately without a restart and the change persists across restarts.

Peak/valley electricity-time indicator beside the Session Log in the DSH Web header. Red light during peak hours (09:00-12:00 & 14:00-18:00 local time), green otherwise.

DeepSeek Harness plugin that steers reasoning toward Simplified Chinese and translates eligible reasoning blocks.

Connects ChatGPT accounts through OAuth, switches among them manually or automatically, and adds Codex models with usage limits to DeepSeek Harness.

Replay real DeepSeek Harness turns against Standard, Minimal, Anchored, or plugin candidates with frozen request-surface evidence

DeepSeek Harness plugin: direct multimodal image transfer declarations + per-session image clearing (surface replace), with a settings page and composer controls.

A desktop-native pet bubble driven by DeepSeek Harness session events: transparent always-on-top Electron window, status-aware animations, balance, pomodoro, roast, and DSH settings card.

A DeepSeek Harness plugin that interviews the user about a project, generates a project timeline / Gantt chart, and exports it as Word or Excel.

Tiered model routing for DeepSeek Harness: think/build model tiers, vision delegation (the vision model only describes images; the working model acts), image generation, and a high-impact guard.

Adds prompt optimization, cancellation, and undo controls to the DeepSeek Harness composer.

Floating workspace file-preview window for the Web client: lazy file tree, CodeMirror preview and in-window edit with save, Markdown and image preview, Quick Open search, and opening paths clicked in the conversation.

Conversation outline plugin for DeepSeek Harness: a side panel and a full-width curtain as two views of the same outline; three search scopes — titles, full text and other sessions; several ways to jump — a heading, a turn header, the end of a section; a full outline or a questions-only reading mode, switchable at any time; a reading position remembered per session; and a Chinese/English interface with a plugin-configuration card.

Message timeline rail for DeepSeek Harness web chat: evenly spaced ticks for each user message on the right edge, click to jump, hover to preview.

Add six languages (繁體中文, 日本語, 한국어, Français, Deutsch, Español) to the DeepSeek Harness Web UI language options, with hand-polished per-language translations and automatic fallbacks (Simplified-to-Traditional conversion for zh-TW, English for the others) covering upstream updates and third-party plugins.

Falsification-driven correction loop for agent sessions: risky work commits an evidence-checkable claim first, deterministic gates verify it, falsified claims force mutually exclusive replacement hypotheses with discriminating experiments, and every event lands in an append-only evidence ledger.

WeChat iLink ClawBot bridge: official Tencent iLink protocol, Web UI settings (QR login / model / preset / permissions / session management with delete), 24h auto-renewal pushes a fresh QR before expiry, zero downtime.

Per-file-type multimodal chains: preset processing models per wildcard convert image/video/audio files into prompt tokens before they reach the text-only session model, with per-preset fallback chains and a Multimodal settings page.

A 轨迹解读 (trajectory interpretation) tab that summarizes each user round — what was wanted, how the assistant thought and executed, files/commands/errors — via a rules engine plus optional LLM narrative; user messages stay verbatim.