Plugins

Self-healing runtime layer: a resource control loop with strategy switching and repair escalation, error-storm and fiber-failure insight, cross-session knowledge carry-over, a self-healing dashboard and 7 tools.

Long-term associative memory for DeepSeek Harness: a local plain-text vault of raw episodes, typed memory cards and entity pages, configured from 设置 → 记忆图谱, and a Memory tab inside a conversation that draws what that session remembered. No build step; the

Daily tarot with 22 Rider-Waite-Smith card images, animated 3D flips, random upright and reversed cards, Chinese readings, and up to 10 draws per day.

Draws every question and answer in the DSH Web conversation blurred until the pointer rests on that row, and locks the mask to always-blurred while the window is unfocused; one control below the composer switches between off, hover to reveal, and always blurred, remembered per browser.

Windows resource monitor for the DeepSeek Harness (dsh) sidebar: samples CPU, memory and clock speed once a minute, keeps a 30-day history and draws separate 24h / 7d charts for CPU and memory (with 85% / 90% alert lines) plus a frequency chart that marks load-time throttling. Reports sustained CPU >= 85% (10 min), memory >= 90% (10 min) and frequency <= 70% of maximum clock under load (5 min).

Kiro (AWS CodeWhisperer/Amazon Q) provider route for DeepSeek Harness, plus an OpenAI/Anthropic reverse proxy. Reuses the existing Kiro IDE/CLI sign-in.

Draws the sidebar navigation as a compact two-column grid at every viewport; on narrow, short or coarse-pointer screens it additionally caps the navigation at 34vh with scrolling so the workspace and session history keep a 38vh floor. Also repairs dsh-resource:// records whose protocol the browser URL parser reports as empty.

Adds a Restart button to the DSH Desktop sidebar. It opens the same native confirmation dialog as the tray Exit item and, once approved, shuts the app down gracefully and relaunches it. This plugin only draws the button: it also needs the companion shell patch shipped in shell-patch/ in the same repository, which has to be applied from outside DSH while the app is closed. Installed on its own, the button stays disabled and a tooltip explains why. The companion shell patch is Windows only and rewrites the installed app.asar.

Replaces detected secrets (OpenAI, AWS, GitHub, PEM, Bearer) with deterministic, structurally valid honey tokens before they reach the model or the session file; adds manual marking for values the detectors miss, explicit user-visible restore, a cross-session canary alarm, and an opt-in JSONL log of substitution counts that never records the underlying values.

Firefly "Spring Missive" theme skin for the DeepSeek Harness web GUI. Overrides 74 --dsw-alias-* design tokens in matched light and dark sets, draws a spring-window backdrop with drifting petals and fireflies, places one character illustration on each side of the layout, and keeps the centre column legible with a radial spotlight scrim. Registers a Settings section with switches for the skin and for the artwork; both choices persist in localStorage.

Two games inside DeepSeek Harness where the model supplies intent and a local engine owns the rules: a twin-brain snake that periodically draws a strategy card, and Chinese chess where the model only comments on a move picked from locally generated legal candidates. Both keep running when the model is slow, offline or out of quota.

Adds a four-pointed-star button left of the composer's model selector that rewrites the unsent draft in place — grammar, wording, punctuation and execution order — preserving code, commands, paths, identifiers, numbers and URLs byte-for-byte, answering nothing and never overwriting text typed while it runs. A pencil under every user message reopens that message for editing without touching the draft; because the Harness session log is append-only, an edit forks the conversation before that message rather than rewriting it. A circular restart icon under every reply asks that one question again in a fresh conversation that carries none of the earlier turns. Both are refused while a turn is still answering, and disabled when the reply used a tool that can leave a trace or when the files the message named have since changed. A right-sidebar memory tree draws the conversation as turns, showing what was actually asked and answered, every answer to one question on its own line labelled original, edited or memory-free, newest first, with arrows that re-read an answer in place; right-clicking jumps to a branch or removes it from the plugin's own history.

Turns the composer's reasoning-effort menu into a draggable slider and takes over the stock model-and-reasoning-level chip. Dragging follows the pointer continuously; releasing snaps to the nearest level and submits one session update per drag. The track draws its own fill, tick marks and white knob, only the highest levels (xhigh, max, ultra) get a flowing brand-colour gradient, and both ends carry the lowest and highest level names. The same card expands into the official model list. Arrow keys step one level, PageUp and PageDown two, Home and End jump to the ends, and each move submits. A generation counter on the directory object discards late responses, and reduced-motion turns the animations off. All data comes from the stock modelDirectories directory; no new capabilities are added.

A small pet in the corner of the interface that reads the real event stream instead of guessing: it drifts when idle, speeds up while the agent is thinking, draws a flowing line per running tool call, turns red and shakes with the error summary on a failure, and celebrates a turn that ran longer than ten seconds. The same widget is the at-a-glance busy and error indicator, so there is no spinner to stare at. The host half folds real harness events into a snapshot and the client half only draws it; it performs no content recognition, and it positions itself away from the composer so it never sits over what you are typing.

DeepSeek Harness (DSH) host plugin: the model draws clean vector figures inline between the paragraphs of its reply, so answers arrive as text-figure-text instead of a wall of text.

DeepSeek Harness 插件：把工作区文件备份到 S3 兼容对象存储（RustFS / 阿里云 OSS / AWS S3）。AK/SK 只从环境变量读取，绝不落盘、绝不进日志。

Tiered compaction hub: replaces the single 80% compaction cliff with a configurable ladder (default 30/50/70/90) whose tiers escalate from gentle to maximum consolidation laws, folds a deterministic keyword index into every checkpoint so facts accumulate across cycles, and summarizes through a model chain — primary plus up to 3 fallbacks, each with its own reasoning effort. Adds a Compaction module sub-tab and a Settings card; toggling off always restores stock compaction.

Live progress for long-running DSH tasks: a script appends JSON lines to a per-session progress file, the Web UI draws them in a floating overlay and a right-sidebar tab that also lists background jobs which reported nothing, and `dsh-progress run -- <cmd>` wraps a command that cannot report for itself.

Redraws line-level diffs on edit, write, insert and str_replace_editor tool cards — unchanged lines render once and changed characters are underlined — covers run_code (PTC) sub-calls, adds a per-turn changed-files card, and opens or copies any of those paths from the card rows and the tool rows through a host route the plugin serves itself.

Tiered compaction hub: replaces the single 80% compaction cliff with a configurable ladder (default 30/50/70/90) whose tiers escalate from gentle to maximum consolidation laws, folds a deterministic keyword index into every checkpoint so facts accumulate across cycles, and summarizes through a model chain — primary plus up to 3 fallbacks, each with its own reasoning effort. Adds a Compaction module sub-tab and a Settings card; toggling off always restores stock compaction.

Local eyes for text-only models: eyes_render draws text/shapes/Mermaid onto a canvas in the Web GUI, eyes_paste captures pasted images, eyes_ocr reads text via the built-in Windows OCR (offline), and eyes_analyze inspects pixels as structured data — no vision model required.