Plugins

完美适配 Purge 的 NSFW content creation 插件。在 DeepSeek Harness 中用自托管模型创作成人向图像、有声视频与多镜头短片，支持参考图编辑、LoRA 风格控制和 Agent 分镜剪辑。

Qwen Image 2.1 workspace for DSH: Apple Silicon MLX or your own NVIDIA server, native reference editing, LoRA, ControlNet, inpainting, shared presets, gallery and live agent progress. Includes verified updates and an agent setup guide.

Session-scoped GeoGebra canvas with 20 native tools for drawing, scene readback, text layout checks and animation capture, plus 55 mathematical skills and bundled offline applet assets.

Local meeting transcription for DeepSeek Harness on Apple Silicon macOS: dual-track recording, live transcript drafts with speaker labels, and @-referenced summaries into decisions and action items.

Native Rapid-MLX provider for DeepSeek Harness (dsh) — dsh reads model facts from the server instead of your settings.yaml.

Vision-first Mac computer use, for when the accessibility tree is unavailable. Ships 11 desktop_* tools - window list, screenshots, OCR text targeting, numbered elements, coordinate clicks, typing, scrolling, drag - plus a Codex-style cursor overlay with spring animation. Clicks are delivered to the target process, so the real pointer never moves and nothing steals focus. In DSH it shows up as Mac Desktop Control with a settings page that checks the engine and the Screen Recording / Accessibility permissions, with a button that opens the matching System Settings pane. Works out of the box: the tarball bundles the dsh-cu engine and the overlay as prebuilt universal binaries (Apple silicon + Intel, macOS 11+, ad-hoc signed), compiled from the repo's Swift sources by the release workflow - no clone, no Xcode, no install.sh. The only setup is granting DeepSeek Harness Screen Recording and Accessibility once. macOS only; on Windows and Linux the plugin loads without registering the desktop tools.

WeChat personal assistant that manages Apple Calendar and Reminders with confirmation before writes and sends scheduled WeChat messages, for macOS 14+ on Apple Silicon.

Xcode AI integration for DSH: 26 Xcode MCP tools (mcpbridge) + Apple platform skills + Xcode Intelligence-style persona (agent preset or global bundle).

Layered long-term memory for DeepSeek Harness - a merge of AirPineApplea/feiyu-memory 4.0.2, dsh-soul-md and dsh-shallow-memory (do not install it alongside those plugins): a diary distilled from finished turns, a hand-written feelings diary, a persona card with version history, a hand-note injected in full every turn, and a short-TTL note layer, each capped separately so per-turn injection stays small; ships agent tools and a sidebar panel for recall, editing, compaction, and backup or restore from a local file, WebDAV or a gzip archive.

Local-first, full-duplex voice for DeepSeek Harness, orchestrated by Muxiva

One-step local model onboarding plus a light harness for DSH. On load it probes loopback and private-network inference endpoints (llama.cpp, Ollama, vLLM, LM Studio), reads each model's context window and output budget from the backend, and registers the provider through settings, so a local model becomes selectable without hand-writing config. It then trims only that provider: masks the tool list through tools.restrict, drops named system-prompt sections and memory contexts, swaps the injected instruction block for a short one while preserving its first 80 and last 96 characters, and truncates tool-schema descriptions without touching structure or required fields. Cloud providers are returned as the same assembly object, untouched. The injected block is re-read from disk on every assembly, so editing the file takes effect on the next request with no reload. The panel in the plugin detail page is a live operations view: it shows whether the takeover is active and by which hook, how many characters the last assembly was trimmed from and to, and which providers are online (probed concurrently, with a real one-shot completion reporting latency and reply text per provider). If a local backend is not running, the panel offers a one-click launch: it probes the port first (never starting a duplicate), then starts the matching server (ollama serve, or a custom command you supply) detached and polls until the endpoint answers, clamping the wait to 3-600 seconds, reaping the child it just spawned if it never answers, and naming the paths it searched when it cannot find the binary. A stop button kills backends this plugin started (SIGTERM, then SIGKILL) and refuses to touch a server somebody else launched. When a model has just been downloaded and no backend exists yet, the panel can also scan the machine for .gguf files with hard caps (300 models, 20000 directory entries, Downloads only one level deep), start llama-server on the chosen one (skipping the .mmproj. vision projection and reusing a server that already answers), register the provider under the model id the backend reports, and run one real completion to prove it works. Before spawning anything it does a fit check that counts only the part of the file the GPU cannot hold: hostBytes = fileSize - VRAM, compared against max(total RAM - 2GB, free RAM). A file far larger than RAM is refused outright; a heavy host-side remainder is refused unless you confirm; the same remainder under mmap is only a warning, because the weights land in reclaimable page cache. A load-mode selector (auto / mmap / none / mlock / mmap+mlock) maps to llama.cpp --load-mode, and the default is whatever the engine picks. From 0.4.2 everything above is derived from the platform instead of assumed: home and data directories come from os.homedir() and the OS conventions (LOCALAPPDATA on Windows, ~/Library/Application Support on macOS, XDG_DATA_HOME elsewhere); backend binaries are looked up in the official install locations of each OS and can be pinned with OLLAMA_BIN, LLAMA_SERVER_BIN, VLLM_BIN or LMSTUDIO_BIN; model discovery walks DSH_LOCAL_MODELS_DIR, ~/models, ~/llama.cpp/models, ~/.cache/llama.cpp and the LM Studio data directory, with the WorkBuddy layout demoted to one candidate among them. VRAM is asked from nvidia-smi, then rocm-smi, then sysctl hw.memsize on Apple Silicon (flagged unified, so the same bytes are not charged twice), and any flip to a unified-memory budget is handled explicitly. Before building a command line the plugin runs llama-server --help once and keeps only the flags that build actually has, so older engines get --n-gpu-layers 99 and no KV quantization while newer ones get auto plus q8_0; anything it cannot probe is passed through rather than dropped. LM Studio is started with lms server start --port N rather than being lumped in with the engines that need a model argument, and vLLM gets its own --model example instead of llama.cpp syntax. Windows, macOS and Linux are all exercised in CI, including a windows-latest and a macos-latest job, because a cross-platform claim that has only ever run on one OS is not a claim. Two of those suites exist specifically to catch this machine's blind spots: one replays the whole suite with process.platform rewritten to each OS, and one deletes every advantage this development box has — no nvidia-smi, no rocm-smi, no sysctl, an empty PATH, a home directory with no models — and requires graceful degradation rather than a crash. That last one caught a real class of bug: several assertions hardcoded 'lms.exe', which cannot pass where the executable has no suffix. Measured on the reference machine: a 16.9GB 27.2B MoE with 14.5GB of VRAM holds a 285-691MB working set and sustains 76 tok/s after a 1000-2000 token warm-up, so file size alone is not a fit signal. From 0.4.3 the scanned models are shown as a table rather than a dropdown (one row per gguf, with its size and any fit warning), so you can see what is on disk before choosing. From 0.4.4 the delete button on the provider table removes the provider entry, not the weights file on disk. That distinction matters: what people actually want to clear out is the dead entry - one pointing at a port that never answered, one registered twice for the same port, one whose model file was deleted long ago - not a 15.8GB gguf. Deleting a config is undone by importing again; deleting a gguf is not. The removal goes through settings.mutate with an unset op on \['providers', id], because update merges layers: writing null there leaves a null provider key that the loader reads as a broken config rather than an absent entry, and only mutate/unset is implemented as Reflect.deleteProperty. The result is read back afterwards, because unset on a key that comes from a lower layer (the profile cordis.patch.yml or a home patch) does not delete it - it restores the inherited value, so the call succeeds while the entry stays. When the entry is still there the panel says so and points at the file to edit instead of reporting success, and a DSH whose settings service predates mutate() is told plainly rather than silently ignored. Deleting takes two clicks and clearing the confirmation is as easy as clicking another row; on success the online and baseURL caches are cleared and the tool surface is recomputed, since the provider list changed. The gguf itself is never touched by this plugin - the scanned-model table is display and selection only.VRAM reporting is driver-dependent. From 0.4.5 model discovery also reads Ollama's own storage: its blobs are named sha256-… with no extension, so extension matching could never list them. Weights are now accepted by extension OR by GGUF magic header (only dot-free files of at least 8MB, capped at 600 header reads per scan; obviously-not-weights extensions like .json/.md/.safetensors/.part are never opened). Ollama models are resolved from their manifests: the model layer is picked by mediaType application/vnd.ollama.image.model (never adapter/mmproj layers), the human-readable name is restored from the manifest path (…/qwen3/8b becomes qwen3:8b, …/latest stays bare), tags sharing one blob merge into a single row, and the metadata kv supplies name, quantization and context length. Adopting an Ollama model does not start a duplicate llama-server over the same weights: it confirms ollama serve answers on 11434 (launching it if not) and registers the provider under the tag the endpoint itself reports — the blob's sha256 name would never match anything, since 11434 lists qwen3:14b, not sha256-a8cc…. Two real-machine-only traps fixed on the way: existsSync('<file>/') returns true on Windows, so manifest-vs-directory branching now goes by parseable content instead of trailing-slash semantics; and the quantization label table had been copied from a stale ggml enum generation, labelling ggml_type 15 as Q5_K_S while ollama show reports Q4_K_M. Same-blob dedup is by (dev,ino), keeping whichever entry has a name. Verified on the reference machine: the scan goes from 1 row to 4 — qwen3:14b (8.64G) and qwen3:8b (4.87G) now appear with Q4_K_M badges — in 4ms; npm run verify exits 0 (launcher 57, portable 35, panel-host green including the linux/darwin simulated-platform runs); the new panel-host suite builds a real miniature Ollama store in tmpdir and points OLLAMA_MODELS at it, so results no longer depend on whether the test machine happens to have Ollama installed. From 0.4.6 the panel shows each model its real name instead of its file name, because some download tooling stores every weight under one generic file name, leaving the real name only inside the GGUF header. The scan now reads each file GGUF KV section - not a fixed-size head buffer, because tokenizer.ggml.tokens (hundreds of thousands of strings on a large MoE model) and the chat template push general.file_type past any window; and GGUF string arrays are interleaved (u64 length plus payload per element), so treating them as a length table walks the cursor off the rails and produces astronomically large reads. Reading and skipping are accounted separately: key names and types are always read, string values over 256KB and string arrays are only walked past without copying, non-string arrays over 64 elements are skipped outright, and a cursor past the file size stops the read, because an out-of-range position means the cursor is wrong rather than that a limit should be raised. The displayed name goes Ollama tag, then general.basename plus size_label, then general.name, then the file name, with the full string kept for the hover text since it usually carries internal build numbers; mmproj files are now identified by general.type as well as by name. A short read (truncated file) no longer feeds half a value into the parser. On the reference machine, reading one file header parses 44 keys in about 16ms, and a whole scan takes about 31ms. From 0.4.7 the smoke test no longer fails the first time and passes the second. The old code gave both request attempts one shared AbortController and one shared 30-second timer, so the total budget covered both rounds: a cold load of a large model burned it in round one, the abort left the signal permanently aborted, and the second round (the Ollama-style reasoning_effort: none retry) could not even be sent - which is exactly why the first click returned This operation was aborted and the second succeeded. Now each attempt gets its own controller and timer, a timeout retries once automatically (retryOnTimeout, on by default) because a first-request timeout almost always means the weights are still being read from disk, and non-timeout errors (refused connection, HTTP error) are not retried at all. Timeouts report themselves: the message says the wait exceeded smokeTimeout, whether a retry already happened, and points at the new smokeTimeout setting (seconds, default 60, clamped to 5-600) - port responding is not the same as tokens coming out, so it is a separate knob from launchTimeout. Unrelated: an example naming one specific model was dropped from the panel copy, since not every machine has that model downloaded, and the explanation now describes the mechanism instead. From 0.4.8 adopting a plain .gguf no longer registers an unrelated Ollama model. The port field used to follow the last click rather than the file: selecting an Ollama entry moved it to 11434, selecting a plain .gguf did not move it back, so a 11434 left over from an earlier click stayed. The server accepted it as given, probed the port, found Ollama answering and borrowed it (the skipped-launch log line about the port already answering is this path), and then the model id - derived from the file name - could not be matched against Ollama's /v1/models, where the fallback was to grab the first entry in the list. Embedding models tend to sort first, so the adopt registered a model the user never selected. Now the port follows the file: a plain .gguf landing on another engine's port (ollama 11434, vLLM 8000, LM Studio 1234) is moved to 8090 and the panel says so. Only ports belonging to a known engine are overridden - a hand-typed 9999 aimed at your own llama-server is left alone. The model-id fallback no longer guesses either: it takes the only entry when there is exactly one (a single-model backend may legitimately report a different id than the file name) and otherwise reports an error listing the ids it did see, rather than registering a wrong provider. From 0.4.9 the panel can no longer vanish. It used to be possible for the whole section to disappear from the plugin detail page while the host stayed up, the HTTP bridge kept answering, and no crash log appeared. The cause was a scope error: a helper was declared inside one component and referenced from another, which is valid JavaScript and passes a syntax check, so it only threw at render time - when the user pressed the scan button and React ran the state updater. The host wraps every slot entry in an error boundary whose fallback is an empty anchor, so the throw was contained silently: the panel was not uninstalled or crashed, it was replaced by an empty div. The helpers now live in the module scope both components share. The client half also gained its own test - a minimal React stand-in loads the real panel module, really calls the components, really runs the state updater at render time, and presses the scan button, so this class of error can no longer reach a release.

A liquid-glass skin built on your current desktop wallpaper: the wallpaper is blurred and veiled into the window canvas, panels stay translucent, and every surface that carries text (composer, code blocks, preview pane, menus) gets its own frost level so nothing is washed out. Apple-style liquid-glass controls. On Windows the canvas reads the current wallpaper from APPDATA; other platforms fall back to a painted gradient.

Native Anki workspace backed by one real headless Anki collection: browse, edit, add and review notes, suspend cards, import and export APKG, and sync with AnkiWeb, with the same collection exposed to the anki AI tool. No Anki desktop app, AnkiConnect or Qt is started. Supports Windows x64 and macOS 12 or later (Apple Silicon and Intel). The matching headless Anki core, CPython and all Python runtime dependencies are bundled, so no separate Anki or Python installation is required.

PDF outline navigation and local reading bookmarks in the existing right-sidebar file preview; supports adding, renaming, deleting and revisiting page positions, with version and integrity guards. Verified on macOS Apple Silicon with dsh Desktop 0.2.1-rc.2 and 0.2.1-rc.3.

Read and write Apple iCloud Calendar over CalDAV through five tools: list calendars, read a window with recurring events expanded, create an event, and delete behind a dry run that must confirm the exact object count. Credentials come from DPAPI, Keychain or libsecret, or from an environment variable.

Ink-and-paper Apple-inspired theme for the DSH web client with a handwritten jimmy sidebar brand.

Find photo candidates from the 2026 Qiandao Lake cycling race by bib number and visible gear, with local helper scripts and album snapshot comparison. Needs the event's photo album link; snapshot comparison uses Python 3.10+, bib OCR is macOS only (Swift and Apple Vision), and the optional gear ranking needs Pillow, PyTorch, Transformers and a local DINOv2 model directory.

A desktop pet for DeepSeek Harness: an always-on-top character and a roster panel that show what the watched conversations are doing, with seven animations, Workspace grouping, subagents drawn under their parent, and one click to open a session. Ships Windows, macOS (Intel and Apple silicon) and Linux window binaries in the package, so installing runs no install scripts and needs no build step.

Keeps the Mac awake for as long as the DSH host is running, whether or not an agent is working: starts ClamAwake's virtual display (mirroring the main screen) plus caffeinate -d, stops both when DSH exits and restarts the engine if it crashes. Self-contained (bundles the arm64 vd-engine, GPL-3.0). Requires macOS on Apple Silicon.

macOS (Apple silicon) fork of aureth-labs/dsh-aurabot-qq: the same QQ persona bot (chat, voice notes, stickers, memory, song covers, modules) with the Windows-only code removed. It connects to a OneBot v11 gateway such as SnowLuma running in OrbStack or Docker, falls back to the macOS `say` voice, sends files over 30 MB to a containerised gateway through a loopback-only temporary download link, and installs the song-cover environment (Applio, audio-separator, uv, Rubber Band, separation models) at pinned versions. It keeps the upstream package name and plugin id (dsh-aurabot-qq), so it replaces upstream and cannot be installed alongside it; use upstream on Windows.

Personal reading log panel: track books as want to read, reading or finished with dates, page progress, ratings, tags, notes and excerpts; yearly reading stats; covers looked up from Apple Books, Google Books and Open Library; Goodreads and Douban CSV import and CSV export; and five agent tools: bookshelf_list, bookshelf_stats, bookshelf_add, bookshelf_update and bookshelf_add_excerpt.

DSH sidebar container panel for macOS Docker runtimes and Apple container

Native macOS tools for DeepSeek Harness agents, with 21 local tools for notifications, clipboard, screenshots, speech, keep-awake, dark mode, Spotlight, Shortcuts, AppleScript and Safari tabs. Zero dependencies.

Editable PowerPoint authoring workflow (a DSH-skill adaptation of icip-cas/PPTAgent) with concise Chinese business styling, HTML slide rendering and PPTX export review. Linux/WSL or macOS (Apple Silicon only); install LibreOffice, Chinese fonts, Node.js 22+ and uv first, then run the one-time setup, which creates a uv-managed Python 3.12 environment and installs Playwright Chromium.