Plugins
Browse, filter, and install DeepSeek-Harness plugins.
1044 plugins found
dsh-selection-explain
yfwu2020/dsh-selection-explain
Select text in the DSH web UI to stream a translation and an explanation of what it means in that specific context, read against the conversation it came from; the panel keeps a replayable history and can be promoted into a full session.
dsh-model-router
neptune810/dsh-model-router
Keeps a pool of models and routes every agent step to one of them with its reasoning effort set — configured by clicking in the composer, no JSON needed. The router takes over the composer's model cell, so one control shows the session's model and reasoning effort alongside the router's own settings. Three authorization scopes (model / effort / full) where a manual pick only counts in the scope that owns it, and in effort scope the router is bound to the one model you choose. Task presets with per-model weights, vision routing for image steps, session signals (active todo, context pressure, token budget, subagent depth), and an optional semantic task classifier. Thinking stays on, because DeepSeek rejects a thinking-enabled request whose history contains a tool call made with thinking off.
dsh-phoenix
stvli/dsh-phoenix
Graceful restart + client auto-reconnect + cross-restart goal continuation for DeepSeek Harness (dsh). After a plugin update dsh-phoenix restarts dsh at a safe idle boundary, keeps the browser page connected, and resumes a running goal so a task can keep
dsh-dupguard
zqh260619/dsh-dupguard
Stops a streaming model call when the same string repeats ten times in a row (threshold configurable), keeps the partial answer as a normal assistant message, and shows a stop notice with the workspace, session name and the repeated string. Repeat counts can vary by unit length through a piecewise table or a JavaScript module, and code regions (fenced, inline, indented) use a relaxed threshold.
dsh-bgjobs
bitsmug/dsh-bgjobs
Runs commands as background jobs that keep executing when DSH exits, with an optional sandbox to constrain file effects, plus a live web panel and offline CLI/GUI management. Windows only for now.
agentduel-dsh
linconz/agentduel-dsh
AgentDuel is a code-battle game. This plugin lets you complete the entire game loop inside DSH: manage deathmatch characters and capture-the-flag teams, submit agent code, start battles, watch replays, and use DSH conversations to keep improving agents based on battle results.
dsh-mindmap
guhanfei-ai/dsh-mindmap
Conversational Markdown mindmap workspace: edit complete documents step by step, keep a live right-side visualization synchronized, and open, manage, and export mindmaps from the working directory.
dsh-vision-hub (bridge-preview)
xing666173/dsh-vision-hub
Inline preview for \[图片: path] image bridge markers: after the marker renders, the image shows in chat automatically, keeping long instructions out of the conversation.
dsh-punky-swarm (dsh-punky-swarm)
punky971210/dsh-punky-swarm
Engine-enforced guardrails for DeepSeek Harness agent swarms: quality gates reject half-done work, crash-safe checkpoints resume in place — keeps AI teams from breaking, not just running.
dsh-plugin-heartbeat
littleblacktong/dsh-plugin-heartbeat
Autonomic heartbeat for each root agent: silent pulses diff the session's todos, Agent Team board and goal against what the last pulse saw, and speak only on a signal — L1 report, L2 ask (ask_user_question), L4 escalate — never starting work. Each message carries a why-now line, an optional one-line physiology signature (time since your last message and how many times it has reached out today) and a compact task snapshot (800-char cap); a dedupe window, quiet hours, a daily budget, refractory/restart cooldowns and persisted vitals keep it quiet, while busy-safe followup injection never stacks. It shares one advisory speech floor with dsh-plugin-memory (~/.dsh/proactive-cooldown.json, 180 s) so the two never talk over each other, stays silent while a child subagent or teammate is still working, and surfaces a "subagents working · N" pill in the conversation header (polling pauses while the tab is hidden). v0.8 adds a diagnostics panel that explains recent quiet pulses and names the current speech-floor holder, plus panel controls for the arbitration window, presence suppression and the physiology line; its settings section carries a stroked heart mark (16×16, stroke 1.3, same spec as dsh-plugin-memory). v0.8.1 makes the plugin compatible with DSH 0.2.2-rc.2: the agent registry is read with agents.roots() when available and falls back to agents.list() on rc.2, and every agent/created listener is exception-guarded — on rc.2 an unguarded throw there rejects agent registration and no session can be created, so rc.2 users must upgrade to 0.8.1 or newer.
dsh-result-only-view
yeyeyeyeshifu/dsh-result-only-view
Results-only conversation toggle for the Web GUI: hides thinking and tool-call process rows, keeps one live status line while the agent works, restores working animations under prefers-reduced-motion, zh-CN/en localized. Works in the stock composer and inside dsh-split-screen panes; verified on DSH 0.1.1 and 0.1.5.
dsh-plugin-chat-mode
xiaofeng7418/dsh-plugin-chat-mode
Adds a tool-free Chat preset for everyday conversation: no tool schemas are sent, the system prompt is one short tone line or empty, and no OS, git or file-tree snapshot is injected. Measured on a plugin-heavy install, one message costs about 1.9K instead of 36K input tokens (roughly 19x fewer); thinking starts off unless you turn it on; on-device DSH Go verified; v0.2.0 adds local slash commands such as `/chat`, `/tone`, `/temp`, `/web` and `/think`, plus opt-in web search that only costs model tokens; v0.3.0 adds web-like composer toggles for thinking and web search that stay in sync with those commands; v0.3.1 gives them DeepSeek-style SVG icons, hover hints and motion, and keeps unrelated global tools out of the request; v0.3.2 adds the web app one-shot icon rotation, removes the square host focus frame, and quiets the hint after a click.
spine-plugin (dsh)
ghabix/spine-plugin
Subtask-tree context manager for DeepSeek Harness (@spinejit/dsh-spinejit): adds spine_open, spine_close, spine_next and spine_spawn tools so the agent works on a persistent SpineTree, replaces closed subtasks with their submitted memories while keeping active work and earlier prefix tokens intact, and renders the live tree in the Web/Desktop input dock and dsh-tui status rows.
dsh-link
k-w7/dsh-link
Keep using a desktop DSH session from your phone: chat, watch the live output, and get the result on your phone when a desktop turn finishes.
dsh-connect-doubao-plus
chsw-cc/dsh-connect-doubao-plus
Fork of Wahwayhale/dsh-connect-doubao that connects a Doubao account to DeepSeek Harness as a model provider after one browser QR sign-in. Keeps the upstream menu-driven model discovery, composer billing multipliers, per-model thinking ladders, task-mode models on paid quota, chain-of-thought passthrough, Seedream image generation and read-only quota panel, and adds a daily check-in that can claim the reward and re-claim it on a timer, monthly and incentive quota windows instead of the fixed five-hour and weekly ones, a model picker that offers exactly the models ticked in the settings card, Windows / macOS / Linux session sealing, and stream hardening: a rejected conversation is never reused, the same-turn fallback opens a fresh one, three consecutive upstream failures trip a local circuit valve, and an interrupted stream stops the server-side generation.
dsh-remedy (plugin)
jjjjjjjjnnjnn/dsh-remedy
A patch set for DeepSeek Harness that repairs fused tool-call argument text at the llm/stream seam and re-throws whatever it cannot prove: a call is dispatched only when every fragment parses to a plain object, the first call keeps its id and synthesized ones are marked, and a stream that cannot be split keeps its original error instead of a success. It also listens on agent/request-error to name the deterministic refusals it will not retry, answers per failure class which patch covers it and whether that patch is switched on in the profile you point it at, and reads multi-frame zstd session logs offline to count every turn ending and every error code. Everything it claims is registered with the seam it touches and the limits it does not cross; the resolver never invents a value and never downgrades a configured setting.
dsh-mesh
fishexploi/dsh-mesh
LAN task delegation between DSH hosts on one network: the parent dispatches a task to a child host, the child works in its own session and workspace, results and files return to the parent's inbox, sensitive actions wait for the parent's approval, and a finished task can be continued with one follow-up inside the child's existing session. Since 0.4.0 a fresh host listens on loopback only and keeps its listener off until it is enabled, the six-digit pairing code is single-use and rotates with a lockout after repeated failures, tokens travel in an Authorization header and are stored hashed, and a dispatch with no grant or with a permission above the grant is refused before anything is sent. Transport is still plain HTTP with no TLS, so use it on a trusted LAN.
dsh-archify-projects
guicaiyue/dsh-archify-projects
Architecture map workbench: register local repos, let a restricted agent (run as a hidden subagent session) read the code with the archify skill, keep branch-level structure and domain diagrams with concepts, temporary/permanent custom diagrams, read-only Q&A, and pick the analysis model from the models DSH already has configured.
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-phone-mirror
liyuhan7/dsh-phone-mirror
Mirrors an Android phone or tablet into the DSH right sidebar over adb with mouse and keyboard control, keeping videos, clips and chats next to the desktop instead of in your hand; Opus audio needs Android 11+, Windows only.
dsh-agent-soul
danshuyue/dsh-agent-soul
Adds personality injection, a decaying memory card, a mood state and a daily diary so the agent keeps identity and history across sessions.
dsh-plugins (dsh-cockpit)
romany-saad/dsh-plugins
Keep a multi-PR effort on one board you can trust at a glance: the hub agent keeps a tree of steps, each linked to its issue, PR, branch and worktree, through four board tools, and you see what is done, what is blocked and what needs you in a right-sidebar Board tab and a progress strip above the composer. Each board is one plain JSON file your scripts can read, and an optional drift command you configure flags steps whose status no longer matches their work.
dsh-plugins (dsh-automations)
romany-saad/dsh-plugins
Keep your automation tasks in files you can version and review: write each task's schedule, prompt, model, permission and working folder in YAML, JSON or JS, and DSH's own Schedule is kept exactly in line, creating, updating, disabling and deleting tasks within a second of a save. Each task runs in its own conversation, persistent or fresh per run; /automations syncs on demand and names any file mistake; and a sync that finds no tasks never wipes the Schedule.
dsh-cookie-guard
3150481853/dsh-cookie-guard
Windows-only. Counts the dsh-auth-* login cookies the DSH desktop shell has accumulated, shows how much of the 16 KiB request-header budget they eat and how many launches are left before the shell stops opening with "desktop welcome: Web authentication failed". The web UI gets a draggable floating card that re-checks every 10 minutes, keeps one snapshot per local date and can refresh on demand; a standalone Windows checker ships in desktop/.