Skip to main content

Plugins

Browse, filter, and install DeepSeek-Harness plugins.

190 plugins found

Q

dsh-mem-weave

quaony/dsh-mem-weave

MemWeave — long-term memory and a persona editor for DeepSeek Harness (DSH). Seven memory layers with per-layer half-life and a decay floor feed three capped injection sections (persona, first-turn snapshot, per-turn delta plus recall); recall fuses FTS5/BM25, embeddings and links with RRF. Writes pass source detection, dedupe/merge and conflict freezing, are audited and can be undone for 24 hours; credentials are blocked and sensitive fields redacted before injection. 22 weave_* tools, nine settings pages, promise and anniversary reminders, one local SQLite file, no runtime dependencies.

63 days agoMemoryMIT
F

dsh-web-search

fangchenghao/dsh-web-search

Keyless web search with one web_tools entry and eight actions for multi-query search, date windows, packages, papers, repositories, platforms, page reading, cited research, cache search and engine health; fuses eight engines and ships a stdio MCP server for other agents.

62 days agoTools & CapabilitiesMIT
L

dsh-computer-use

ltmroberthk915/dsh-computer-use

Windows Computer Use for DeepSeek Harness with 21 tools and verified targeting: controls are addressed through their UI Automation identity and patterns, and numbered marks are checked against the live window before input, so a stale target is refused instead of clicked. A known sequence runs in one computer_batch call, one model round trip; incremental observations, compact receipts (normally no more than about 6 KiB extra), bounded handoff waits (HANDOFF_WAIT_TIMEOUT/HANDOFF_HELD_TIMEOUT as continuations), the desktop-block cycle boundary and a deterministic 100 s owner-scoped leave backstop keep context small and prevent a stuck handoff from waiting out the tool deadline. Approval modes settle an unanswerable ask honestly: in a danger-full-access session with no approval card the action runs under that grant, every other sandbox — and every delegated subagent session — refuses and names the unavailable channel instead of reporting a user rejection, and gate diagnostics land in policy.jsonl. Human handoff, a model-visible stop brake and prebuilt native workers.

6yesterdayTools & CapabilitiesMIT
S

specpowers

saxem1997/specpowers

Registers six spec-driven + test-driven workflow skills for DeepSeek Harness: OpenSpec spec authoring fused with Superpowers TDD discipline into a Phase 0-4 pipeline with review gates. Zero runtime dependencies.

6last monthSkillsMIT
L

dsh-nebula-mind

lng-ux/dsh-nebula-mind

Cross-session memory for DeepSeek Harness: three-way recall (BM25, dense vectors, and Personalized PageRank over a memory graph) fused with RRF, emotion as a modulation multiplier, power-law forgetting with five sinking layers, and an interactive memory star map.

522 hours agoMemoryMIT
X

dsh-database-explorer

xing-he888/dsh-database-explorer

Database explorer panel for DeepSeek Harness: connect SQLite, MySQL/MariaDB, PostgreSQL, SQL Server, MongoDB, ClickHouse, Redis, Elasticsearch and Qdrant from a third Conversation tab — browse schemas, page through data, edit cells, insert rows, design and drop tables, run SQL, import/export CSV/JSON/SQL/Markdown, and draw editable E-R diagrams. Read-only connections enforce statement-level allowlists, agent tools ship read-only by default with an opt-in write flag, and every write (including refused attempts) lands in an auditable log. Ships an MCP stdio server exposing the same connections to external AI clients.

5last monthTools & CapabilitiesMIT
S

dsh-wsl-suite

stallenlu/dsh-wsl-suite

WSL workspaces for DeepSeek Harness. A sidebar W button lists installed distros and registers a Linux folder as a host workspace addressed by a UNC path, and the sidebar terminal opens a real Linux pty inside the distribution for WSL sessions. A persistent zero-dependency agent running in the distro answers file, shell and search requests over a JSON-lines pipe, so no polling timers touch the distribution while it is idle and /mnt drives stay reachable from tools and the terminal (registering one as a workspace is refused, because the host realpaths workspace paths and Windows cannot resolve drvfs); grep and glob run the distro's own ripgrep. WSL preset twins such as wsl-standard swap in the WSL shell and fs tools. A rebuild of 6Mikao9/dsh-wsl-workspace — do not install both: they mount the same W entry in the Workspaces header, publish the same wsl-<mode> presets, and each wrap the host's single-slot terminal provider.

45 days agoDev & Plugin ToolsMIT
A

dsh-bili-interactor

acttk/dsh-bili-interactor

Browses Bilibili on your behalf and actually takes part: search videos, like, coin, send danmaku, post comments, reply to other people, and keep replying when someone answers back. Reads the video's Bilibili AI subtitle (or transcribes the audio locally when there is none) so its comments talk about what the video really says, and it reads that same content before choosing whom to reply to. Instead of copying other people's sentences, it distills rules about how to sound like a person and gets less AI-ish the more it runs. Skips videos it already browsed, de-duplicates replies per thread, and after too many rounds with the same person in one comment section it reviews whether the exchange has any substance -- if it looks like two bots complimenting each other, it says one last line and stops replying to that person there. Checks whether each comment is genuinely public or only visible to your own account, can scan any time window (last hour / day / 3 days / all) and lists the ones others cannot see, warning you on the overview when replies got folded. Six-tab settings panel where every change takes effect immediately with no restart, a manual run that queues behind the background listener instead of being refused, and an environment self-check that tells you exactly what to install. Requires Python 3 for the Bilibili actions; the built-in check lists any missing packages.

424 days agoTools & CapabilitiesMIT
C

dsh-subagent-hub (subagent-hub)

ccsert/dsh-subagent-hub

Named subagent presets bound to model-route allowlists, where an out-of-list child model is refused before a delegation starts and each child conversation shows the model its last request actually used.

48 days agoModels & ProvidersMIT
H

dsh-discipline-guard

haozheou/dsh-discipline-guard

Agent discipline guard for DeepSeek Harness: four gates enforced at the harness boundary — loop circuit-breaker, token-cost fuse, route-switch warning, plan approval — where prompt-level reminders can't reach, so an agent can no longer burn a weekly quota in minutes; one-click 🛡 chip, hot-reloadable thresholds, and presets from invisible-but-armored to warn-only.

43 months agoUsage & BillingMIT
F

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.

38 days agoWorkflow & AutomationMIT
C

dsh-oracle-gate

cresbigb/dsh-oracle-gate

Acceptance gate for DSH: freeze the acceptance criteria into an oracle ledger before work starts, then honor only real execution receipts — files changed after freezing are refused outright, and a failing hard gate overrides any human pass.

316 days agoDev & Plugin ToolsMIT
L

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.

33 days agoModels & ProvidersMIT
Q

dsh-web-search-fetch

qikairo7/dsh-web-search-fetch

Aggregated web_search and web_fetch providers for DeepSeek Harness, sitting behind the built-in tools. Search fans one query out to AnySearch, Zhipu and Zhihu concurrently, dedupes URLs, fuses ranks with Reciprocal Rank Fusion and optionally reranks through SiliconFlow. Fetch is URL-aware: static pages pass through the built-in HTTP fetcher, dynamic ones fan out to Firecrawl and AnySearch extract and pick one body by quality. One shared circuit breaker skips an engine for the day after a quota or 429 signal, and a settings card keeps each engine API key in the credential store without ever showing its value.

38 days agoTools & CapabilitiesMIT
A

dsh-plugin-memfuse

a1141171521/dsh-plugin-memfuse

Long-term memory for DSH stored as plain Markdown. Real user/assistant exchanges are captured into daily session files without a model call; an LLM pass then distils them into memories that are injected into the system prompt each turn. Retrieval is local (CJK bigram + BM25) in three levels — index, summary, full text — so the model only pays for the detail it asks for. Memories carry priorities; P0 and user-written entries are never changed automatically, and forgetting marks an entry stale instead of deleting it, so the whole chain stays auditable in status.md. Six tools cover recall, write, status, extract, feedback and the skill library; when a cluster of memories is ready to become a skill, a card appears at the end of the turn, writes the SKILL.md on click, and installs it only after you approve. Learning is on by default and follows the conversation model; turning the master switch off stops the LLM entirely while capture keeps working. Failures are surfaced in the prompt and in status.md instead of being swallowed. Pure JavaScript, no third-party runtime dependencies beyond the harness packages.

32 days agoMemoryMIT
S

dsh-session-delete

shuiqian520/dsh-session-delete

Permanently deletes one session from the sidebar session menu, or several at once from the "..." menu beside the sidebar section title: tick sessions in the list (running ones are refused), then confirm in a dialog — log directory, workspace membership, live Host instances and the client list row are all cleaned. The same "..." menu also collects the session search and view options. It also sweeps rows whose log is already gone, and retries a user message or an assistant reply from the button right of Copy.

3last monthSessions & MessagesMIT
Y

dsh-simple-restart

yizhixiaokong/dsh-simple-restart

Adds a restart row to Settings → General that relaunches dsh web with the same invocation through a detached helper, refuses under a detected systemd supervisor unless allowRestart is true, and tests whether the listening port is free by connecting to it rather than binding it.

310 days agoDev & Plugin ToolsMIT
K

dsh-fuse

klrsl/dsh-fuse

Page-level UI generation for DeepSeek Harness: design tokens (theme.json) and code standards (code-style.json) drive a dsh-fuse fence that renders whole pages, with spec pre-validation, a theme contrast guard, a pixel-level style inspector (getComputedStyle), undo history, and a settings UI.

32 months agoUI EnhancementsMIT
P

dsh-bayes-predict

pg527322814/dsh-bayes-predict

Bayesian stock prediction plugin that fuses MACD, RSI, moving-average, volume and other technical signals into an up-move posterior probability with multi-horizon comparison, trend-state classification and position risk metrics for A-share and US stocks.

3last monthTools & CapabilitiesMIT
M

dsh-refuse-compaction

mizuki5101/dsh-refuse-compaction

Intercept auto context compaction before it runs — four tiers (ask first / delay / forbid) with an occupancy gauge; manual /compact untouched. Built for local small-window models.

23 days agoSessions & MessagesMIT
J

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.

220 hours agoDev & Plugin ToolsMIT
S

dsh-windows-computer-use

shinji00222/dsh-windows-computer-use

Windows computer use for DSH built on a native C# engine: text-only DeepSeek models can read UI Automation controls and fill forms in verified batches without screenshots, stale control refs can be re-found by unique identity, CAPTCHAs and verification codes are refused and handed back to you through the DSH approval prompt, and Esc stops the turn at any time. A derived project based on Yu-tao-Li/dsh-computer-use-win; do not install it together with that upstream plugin, since both drive the same desktop.

215 days agoTools & CapabilitiesMIT
C

dsh-context-toolkit

chimneyq/dsh-context-toolkit

Keeps bulk output out of the conversation: ctk_execute, ctk_execute_file and ctk_batch_execute run code in a child process and return only the conclusion, ctk_index and ctk_search file long material into a local FTS5 knowledge base built on node:sqlite, and a routing gate refuses curl and wget before they execute. Ships ctk_stats and ctk_doctor plus a settings page with a live context-occupancy panel. No third-party runtime dependencies.

2last monthTools & CapabilitiesMIT
D

dsh-whale-post (verify)

deepsleepaquarium/dsh-whale-post

Verifies letters signed by other agents: a canonical field order, digest and HMAC checks, and fail-closed handling of missing or unregistered fields. Off by default; when enabled, a letter that fails any check is refused rather than delivered.

2last monthSecurity & PermissionsMIT