플러그인

Failure memory for DeepSeek Harness. Captures tool failures as they happen, classifies them with 16 built-in rules into lessons pairing a symptom with a fix, adds session/daily/weekly summaries, and injects a short briefing into later sessions through systemPrompt context. Ships 5 pitfall_ tools and a web settings page.

Stores the agent's long-term memory as plain Markdown in an Obsidian vault — every memory is a .md file with YAML frontmatter that Obsidian indexes natively, so you can read, edit, link and back up it with tools you already use. Nothing is locked in a database. Ships a sidebar panel (tag filter, sort by recency or importance, edit in place), 7 agent tools, three-tier token-budgeted recall, BM25 ranking (no embeddings, no API key, works offline, CJK bigram tokenization), importance decay with archiving (moved to .archive/, never deleted), and optional GitHub Contents API sync. A note containing a secret is refused rather than redacted, because a redacted memory reads as complete while the important part is gone. Zero runtime dependencies; everything comes from the DSH host.

Local Ollama worker tools for DeepSeek Harness: model inventory, a single unverified ask, and a batch runner that chunks long text, checks every item deterministically, returns a small verbatim sample for review and can optionally write the labelled result inside the session workspace.

Named model routes for subagent delegations, configured from a settings page: each alias binds a subagent provider, model and optional reasoning effort, token cap, agent preset, tags, plus a "when to use" line the model reads to pick it, so a delegation names only the alias. Adds five tools (delegate, list routes, edit config, wait, release), shows the route a child actually ran on in its session header, lets each UI element be switched off, and keeps all configuration in one local JSON file instead of YAML.

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.

WYSIWYG Markdown notes editor in the DSH right sidebar, with tabbed documents, table editing, chat integration, assistant read/write tools, and local file export.

WeChat Reading (WeRead) right-sidebar bookshelf and chapter reader: QR or Cookie login, store search, resume reading, and weread_shelf / weread_search / weread_book / weread_chapter agent tools.

A composer control: shows the current model's context window, the tokens in use and a rough system / tools / messages breakdown, switches the model's reasoning effort in one click (writing reasoningEfforts into the model config when the provider declares none), and edits the model's contextWindow from the same panel so custom-provider models stop silently inheriting the adapter default. Makes no model calls; reads only the official session projections, model directory and config-forms services.

Optional source-preserving chat rendering with selectable themes, offline Uiverse component tools and a report-local evidence workspace. Preview for DSH 0.2.1-rc.2.

당신과 AI의 모든 대화를 별의 바다로 그려 주는 시각적 미리보기입니다.

Harness Web UI의 원격 Linux 서버 관리자입니다. 파일 탐색·편집, 호스트 상태 확인, Docker 컨테이너 시작/중지/재시작, 이미지 확인, Docker 레지스트리 미러 설정, firewalld/ufw 규칙 관리, apt 소스 전환, 개발 도구 감지 및 apt 설치/업그레이드를 지원합니다.

작성창 옆 Git 칩을 눌러 중앙 작업대(변경사항, 기록, 파일, 태그, stash)와 설정의 Git 섹션을 엽니다. 여러 계정의 OAuth 기기 흐름과 PAT를 지원하며 저장소마다 권한(full/read/none)을 설정합니다. 쓰기는 전역 스위치로 제한하고 읽기는 허용 디렉터리 목록에 한정합니다. 대상 remote에 쓰기 권한이 있는 계정이 없으면 push를 거부합니다. diff는 줄 번호, 4가지 레이아웃, 단어 단위 강조, 구문 강조와 함께 바로 미리 볼 수 있고 Ctrl/⌘+F로 검색합니다. Markdown과 이미지를 인라인 렌더링하고 commit 메시지를 검색하거나 스테이징된 diff에서 호스트 LLM으로 생성합니다. Agent 도구 8개, 로컬 HTTP 엔드포인트 51개, 언어 8개를 제공합니다. 토큰은 저장 시 AES-256-GCM으로 암호화됩니다.

DeepSeek Harness용 납품 워크플로 프리셋입니다. 사용자 요청을 원문 그대로 보존하는 에이전트 프로토콜과 도구 7개로 피드백 결정, 승인 기준이 있는 작업 계획, 실제 검증 명령과 종료 코드를 기록하고 프로젝트별 JSONL 파일에 완료 요약을 작성합니다.

macOS용 완전 로컬 푸시 투 토크 음성 입력입니다. 전역 단축키를 누르고 말한 뒤 놓으면 받아쓰기가 커서 위치에 붙여넣어집니다. 로컬 우선 STT(Qwen3-ASR 0.6B 4bit on MLX/Metal 기본, SenseVoice와 mlx-whisper 대체), 외부 API 미사용, launchd 자체 복구 서비스, 8항목 진단 스크립트를 제공합니다. 내장 Skill은 에이전트가 전체 도구 체인을 설치·설정(Karabiner 단축키 연결, launchd, 엔진 전환)하고 문제를 해결하도록 합니다.

read, read_image 및 vision 도구가 파일을 열기 전에 준비합니다. 2048px 또는 8MB를 넘는 이미지는 축소하고(같은 턴에서 1280px로 한 번 재시도), BMP/TIFF를 PNG로 변환하며 바이트를 판별해 잘못 표시된 업로드 형식을 고칩니다. docx/xlsx/pptx, 텍스트 레이어 PDF, csv/tsv, html을 외부 프로그램 없이 Markdown으로 변환합니다.

한 번 설치하면 DSH 사이드바에 PDF 도구 두 개를 추가합니다. 아무 페이지로 이동하고 로컬 렌더링 페이지와 썸네일로 문서 전체를 탐색하거나 DSH 네이티브 미리보기로 PDF를 엽니다. /pdf 명령도 제공합니다.

API 키가 필요 없는 이미지 생성 도구입니다. generate_image(키 불필요, 한 번에 한 호출, 횟수 제한이 있는 속도 제한 재시도)와 list_image_models의 두 도구를 제공합니다.

게시된 로컬 BeatDesign 0.2.5 작업공간을 시작하고 Projects, Canvas, Assets, Timeline용 MCP 도구 30개를 연결합니다. 공유 프로젝트용 Skill과 원격 생성 승인 안내를 포함합니다. Node.js 22.12+와 비어 있는 포트 3020이 필요합니다. 프레임 추출과 타임라인 렌더링에는 PATH의 ffmpeg/ffprobe도 필요하지만 브라우저 내보내기에는 필요하지 않습니다.

ccf_deadlines, ccf_venue, ccf_update 도구, /ccf 명령, 목록을 캘린더로 내보내는 데스크톱 패널로 CCF 학회 논문 마감일을 조회합니다. 날짜는 실시간 동기화(CCFDDL이 주 출처, 두 번째 출처와 교차 확인)에서 가져옵니다. 출처 간 차이가 36시간 이상이면 충돌로 표시합니다. 데이터를 포함하지 않으며 향후 30일은 system prompt에 유지됩니다.

읽기 우선 VPS 운영 도구입니다. systemd 상태 확인 및 작업, 디스크 사용량, Docker 상태, 로그 끝부분, 설정된 백업, 조용한 시간대 일정을 다룹니다.

로컬 Gopeed 다운로더를 위한 에이전트 도구입니다. 다운로드 생성·조회·일시 중지·재개·삭제, Gopeed 프로세스 진단·실행, DSH 백그라운드 작업을 통한 완료 이벤트 수신을 지원합니다. 완료 이벤트는 유휴 에이전트를 깨웁니다. 프로세스 진단과 실행은 Windows 전용이며 다운로드 도구는 Gopeed가 작동하는 모든 환경에서 사용할 수 있습니다.

에이전트 프리셋별 deny-only 도구 차단 목록입니다. 설정에서 도구를 끄면 해당 모드의 도구 스키마에서 빠지고 호출도 거부됩니다. 도구를 켜거나 허용 목록에 추가할 수 없으며 거부하지 않은 도구는 그대로 둡니다. 각 행을 펼쳐 도구 설명을 볼 수 있고 규칙은 프로필 밖의 JSON 파일 하나에 저장됩니다.

DeepSeek Harness용 기억 기능입니다. 기억, 트리, 프로필, 그래프, 받은 편지함의 5개 섹션과 보관할지 버릴지 결정하는 규칙 기반 수집 정책을 제공합니다. 부모 및 자식 청크를 BM25와 벡터 검색으로 함께 찾고 동일한 로컬 파일을 읽고 쓰는 에이전트 도구 12개를 제공합니다.

DSH용 로컬 우선 조사 도구 모음입니다. 모델 도구 kb_grep, kb_list, kb_save를 등록합니다. kb_grep은 ripgrep이 있으면 이를 사용하고 없으면 grep으로 Markdown 지식 베이스를 검색합니다. kb_list와 kb_save는 범위가 제한된 ctx.fs 파일 시스템 서비스를 사용하며 기본 경로는 ~/dsh-kb입니다. local-first-research와 session-memory 두 기술도 포함하지만 플러그인 번들로 설치되지 않으므로 DSH 기술 디렉터리에 직접 복사해야 합니다. 순수 Node ESM이며 실행 시 npm 의존성이 없습니다. 프레임워크 없는 node 테스트 하네스가 포함됩니다. plugin_manager install_bundle로 설치하며 클라이언트 UI는 포함되지 않습니다.