Plugins
Browse, filter, and install DeepSeek-Harness plugins.
278 plugins found
dsh-sciwork
58catmiao/dsh-sciwork
A research library panel and skill set for DSH: project folders, figures with registered provenance, drafts, references, data tables, journal profiles and submissions, plus writing, citation and Origin-plotting discipline for research agents.
dsh-fullscreen-input
al-yichen/dsh-fullscreen-input
Adds a full-screen panel to the DSH composer for writing multi-line text, movable and resizable, sharing one draft with the composer; Enter inserts a newline and only Ctrl+Enter sends, so an IME cannot turn Shift+Enter into a send.
dsh-font-enhancer
loyalchiiina/dsh-font-enhancer
Global font DIY: change size, weight, colour and family at any place in the DSH interface, per element or for the whole window, then hide the floating ball when you are done. No CSS writing and no restart. Adapts DSH kernel 0.2.2-rc.1.
dsh-wechat-mp-studio
funcwei/dsh-wechat-mp-studio
WeChat Official Account content studio: anti-homogenization rotation writing, low-creativity remediation playbook, blessing-image visual baseline, gpt-image pipeline with OCR acceptance, and the measured xiaolvshu draft web API.
dsh-agent-team-auto-orchestrator
loyalchiiina/dsh-agent-team-auto-orchestrator
Host-side orchestrator for DSH agent teams. Every turn it injects an 11-section orchestration methodology into the system prompt and backs it with a counting tool-layer gate, so parallel dispatch, task-board tracking, cross-session handoff, employee-pool reuse and preset routing become the default in any preset and any session instead of something you have to ask for. The gate counts real dispatches (subagent / subagent_fork / spawn_teammate) rather than words: a todo listing N unmarked items only needs the base threshold, items explicitly marked as parallel must be launched in one batch up to a configurable cap, and a genuine single-step task is exempt by writing one marker. Read-only tools (read, glob, grep, context-management tools) are never gated. Team members are woken up and reused before new ones are hired, and subagents are reserved for work that must run under a specific preset. Local-only settings (preset routes, domain keywords, profile name, active tier) live in <DSH_HOME>/team-orchestrator.local.json and never ship inside the package. Status is exposed on a loopback-only HTTP route. Zero telemetry, no hosted service, no account. Adapts DSH kernel 0.2.2-rc.1.
dsh-local-trim
leonlee114514/dsh-local-trim
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.
dsh-incident-journal
qw2007/dsh-incident-journal
Five skills for a reviewable engineering incident journal: record an event only when a counterfactual can be written, quiz yourself from the existing entries, audit the journal for decay, ingest outside material together with its provenance, and rename a layout directory while rewriting every reference to it.
dsh-sketchpad
bmhty123/dsh-sketchpad
Handwriting and diagram whiteboard: sketch on a floating or sidebar canvas and attach the cropped image to your message.
office-ppt
pawpress/office-ppt
Builds presentations by driving the LibreOffice installed on the machine over UNO instead of writing OOXML in JavaScript: `lo_build` opens a real Impress document from structured slides and lets LibreOffice export editable PPTX/ODP plus PDF and per-slide PNG, `lo_render` re-renders an existing PPTX/PPT/ODP, `lo_convert` converts between Office formats, `lo_probe` reports the detected soffice path, version and UNO bridge, and `lo_config` edits the stored settings. Needs a local LibreOffice (verified on 26.8.1.1); it uses the bundled Python, so no system Python is needed, and it keeps one warm soffice process between calls. Fully verified on Windows only; on Linux it falls back to the system python3, which needs python3-uno.
human-writing-skills
whh110112/human-writing-skills
Multilingual writing, chunked long-form collaboration and continuity review, with on-demand DSH delegation discovery. Requires Python 3.9+ on PATH.
dsh-scholar-suite
docsor1212/dsh-scholar-suite
Eight academic skills bundled for one-command install — publication-ready academic figures designed from real data with visual review, AI-text humanization, paper polishing and rewriting, PubMed citation verification (DOI/PMID reconciliation), Chinese citation and medical-book reference checks, and WeChat medical article structuring. Three skills run fully local, five call external services only when used (NCBI E-utilities, DOI resolvers, publishing platforms), declared per skill. Zero telemetry, no install-time scripts. Skills run local Python 3.8+; the academic-figures skill also needs its pip dependencies installed.
astro-ph-researcher
wing-tn/astro-ph-researcher
Astronomy research workbench for DSH and Codex: NADC terms, reference library, writing, scoring, author confirmation, recommended books, FITS/XML formats, and analysis guides.
dsh-skill-center
tuoluosuan/dsh-skill-center
Browse, preview and install public Agent Skills from inside the DSH web GUI, aggregating five sources (claudeskills.info, skillsmp.com, the awesome-claude-skills repository list, the official anthropics/skills repo, and skill packs shipped by installed DSH plugins). Skills install pinned to a commit rather than a branch name, so what you previewed is what lands; the panel reports provenance, whole-tree update checks and a trash instead of deleting. Before writing anything it validates the skill name and frontmatter against the rules the harness silently enforces, lists files that SKILL.md references but the fetch did not bring, flags partial fetches, and offers overwrite, rename or skip when a name is taken.
dsh-living-memoir
yansera/dsh-living-memoir
Cross-session memory kept as a short Markdown living document that is rewritten in full at the end of every turn, written to by model tools, the page, and automatic rewriting alike.
dsh-file-shield
yazzyk/dsh-file-shield
Blocks an agent from reading, searching, writing, or editing chosen files and directories, keeping their contents out of the conversation — including sensitive-word files whose text can make later provider requests fail with a 400. Rules are picked from a file/directory browser on the plugin page.
dsh-decision-split
yuanbaoerer/dsh-decision-split
Decision split view for DSH web UI: a decision summary card in the main window plus the full plan in an independently scrollable side panel, so you can review the whole plan while writing decisions and send them back to the agent with one click.
dsh-humanizer
deep-ios/dsh-humanizer
中文文本人味化 bundle 插件:强制深层改写工作流(十维叙事审计 + 功能路径 + 15 层分布)+ 内容忠实守卫。编辑辅助,非 AI 检测器。
outbreak-report-writing
naliehu/outbreak-report-writing
Report generator for Chinese CDC field epidemiology: turns an incomplete event YAML or case line-list CSV into a first, progress, final, closure or single-case report as Markdown, Word and PDF, marks missing fields as P0/P1/P2 with deadlines, bins the epidemic curve by the disease's incubation period, and writes a raw and a de-identified version in one run.
dsh-novel-fanfic
drizzlegig/dsh-novel-fanfic
Import the original and you have a fanfic reference library of your own: relationships, plot history and world rules always at hand, and the fates, relationships and plot you choose to change recorded as Fic Canon. While writing, keep knowledge, timeline and scene details straight; after writing, check OOC, canon conflicts and continuity. The original is only a reference — your story is yours to decide.
dsh-novel-voice
drizzlegig/dsh-novel-voice
Want AI to write in the style of novels you love, or to develop a voice that is uniquely yours? Creative Style imports novels you like and your own original work, distils their writing patterns into a personal style profile, and lets you blend several sources into one voice. Before you write it compiles a short brief; afterwards it points out clichés, repetition and off-style sentences, and can rewrite specific spans. Off by default, no scores, no model training.
dsh-novel-fade
drizzlegig/dsh-novel-fade
Writing a novel of hundreds of thousands, even millions, of words with AI — but always worried it will forget earlier plot, characters and foreshadowing? Loading piles of old chapters into the context every time is not only costly; it also crowds out the room that should go to actual creation. Fade gives the AI a long-form memory that gradually fades: the recent plot stays sharp, the distant story keeps its key points, and important characters and events can be recalled whenever needed. When a chapter is finished, its memory is organized and its text archived automatically — no need to carry the whole novel around. Cut unnecessary token consumption, keep more context and budget for the chapters ahead, and let AI help you write a longer story.
dsh-style-detector
grq6414-hash/dsh-style-detector
Chinese prose "AI-flavor" detector in the right sidebar: a local rule engine scores and highlights template phrasing, uniform sentence rhythm, vague imagery and repetition, with a separate typo-check page; the score is always computed locally. The UI switches between Chinese and English.
dsh-chapter-whale
thus-zhou/dsh-chapter-whale
Windows x64 DSH workbench for scheduled, unattended novel writing and Fanqie publishing, with configurable tasks and optional free-quota model channels. First start downloads a pinned Python and browser runtime (about 280 MB) from GitHub into %LOCALAPPDATA%\ChapterWhale; publishing signs in to your Fanqie writer account.
dsh-research-suite
xuzhecomeon/dsh-research-suite
Research workflow suite that registers 10 domain skills and 15 tools, curates 450 bundled research skills into 103 capability cards with an auditable per-skill ledger, and takes over research requests automatically through a system-prompt section. Tools cover real R/Python execution (off by default, two-phase preview then confirm), data gates, figure and docx auditing, .docx generation in Chinese-journal style, live PubMed search, field-level DOI verification, and cross-session project memory.