플러그인

LLM-Wiki형 세션 간 Markdown 메모리와 SOUL.md 페르소나. 기본 주입은 첫 사용자 메시지 뒤 활성 세션에만 적용. remember/recall/consolidate/forget. session format v4(공식 DeepSeek Harness 0.1.7+) 및 source kind plugin:memory 필요. v0.13 engines.dsh는 공식 0.2.1-rc.5 계열만 허용하고 구버전은 거부; Review now의 POST /api/memory/control action growth-now는 floor 준수 recap 1회 전송. v0.14 공용 세션 gate(chip, /memory on|off|status, 설정): off는 주입/digest/회상/last_access 중지, 삭제 없음. disabledWorkspaces로 프로젝트 전체를 기본 off로 설정하고 세션 선택은 7일간 유지. v0.15 Basic soul은 SOUL.md 관리 블록만 갱신하고 나머지 보존, BOOTSTRAP 완료, 기존 soul 반영 및 없을 때 생성 알림. checkup 보고서(우선순위 작업 목록).

Windows용 Git Bash 셸 도구: Git for Windows와 번들된 bash를 통해 명령을 실행하며, 타임아웃, 샌드박스, 출력 잘림, 백그라운드 작업을 갖추고 느린 pwsh 도구와 WSL 전용 bash 도구를 대체합니다.

에이전트의 MCP 서버·스킬·도구를 관리 — 컨텍스트 내 실제 상태(로드/잘림/축출)와 세션별 전환 스위치를 보여줍니다.

작업 품질에 영향을 주지 않고 토큰을 적극적으로 절약하는 DeepSeek Harness 플러그인: 도구 결과의 무손실 앞/뒤 잘라내기(기본), grep 결과 접기, 압력 기반 dsh-compaction-basic 위임, 공격적 모드를 위한 LLM 요약 CompactionEngine 서브클래스. 8개 시나리오 합성 스위트에서 입력 토큰 97.6%를 절약하며, 모든 축소는 dsh 이벤트 흐름을 거쳐 '모델 가시 = 기록됨'이 유지됩니다.

TRAE 스타일 대화 롤백 — 턴별 체크포인트(10턴 슬라이딩 윈도우)가 변경된 파일을 롤백하고 모델 컨텍스트를 제자리에서 truncate하며 동일한 세션 ID를 유지합니다. 사람이 시작한 경우에만 작동: /rollback 명령 또는 턴별 웹 버튼(중단된 턴에서도 작동), 영향을 받는 각 파일과 그 동작(restore, recover, delete, skip)을 나열하는 미리보기를 제공합니다.

웹 GUI용 Codex 스타일 `@file` 멘션: 인덱스 잘림 없이 모든 형식(.java, .vue, PDF, 이미지, 확장자 없음, 숨김)의 워크스페이스 파일을 검색하고 참조합니다. dsh-at-file 대체.

자신의 메시지 말풍선에 편집과 삭제를 추가합니다. 편집하면 이후 턴을 잘라 새 텍스트를 다시 보냅니다. 삭제하면 해당 메시지와 이후 턴을 제거합니다. 두 작업 모두 제거된 턴에서 기록된 파일 변경을 선택적으로 되돌릴 수 있으며, 이후 파일이 바뀌었다면 거부합니다. 사이드바에는 대화 하나만 유지됩니다.

"New Session" 버튼에 내장된 프리셋 모드 배지: 클릭하여 다음 세션의 기본 에이전트 프리셋을 선택(긴 이름은 줄임표로 잘려 레이블이 읽기 쉽게 유지됨).

dsh Web GUI 자동 이어쓰기입니다. 로컬 모델이 output-token 한도에 도달해 max-tokens 알림이 나타나면 짧은 이어쓰기 메시지를 자동 대기열에 넣어 응답을 계속합니다. 이중 소스 잘림 감지, 루프 보호, UI 없음.

기본 제공 캐시 적중률을 소수점 셋째 자리까지 표시하고 잘림을 방지하기 위해 Composer 통계 줄의 너비를 넓힙니다.

스킬이 감지되거나 선택되지 않는 이유(잘린 설명, 잘못된 frontmatter, 누락된 리소스 파일)를 진단하고 설치된 플러그인 간 로더 항목 충돌도 확인합니다. 프로필별 기록을 유지해 매 검진을 이전 결과와 비교하고, 반복되는 발견 사항을 근본 원인별로 묶습니다. 너무 긴 설명을 고칠 워크시트를 만들고 스킬 디렉터리 설정을 건드리는 플러그인 패치를 나열합니다.

단일 AGENT.md에서 페르소나와 메모리를 구성합니다. 매 세션의 system prompt에 처음 40,000자를 주입하고 초과분은 안내와 함께 자릅니다. Settings 페이지에서 편집할 수 있습니다. 턴별 일기 항목은 유사도에 따라 병합하고 회상 빈도로 그룹화하며 총 2MB 한도에 도달할 때만 정리합니다.

턴이 비정상 종료되었을 때 자동 복구합니다. 일시 오류 후 재개, 출력 잘림 후 이어쓰기, 할 일 목록 미완료 후 재실행을 지원하며 요청 단위 429 대책과 편집 후 재전송, 수동 재시도, 중지 후 다시 묻기 기능이 있는 클라이언트 도크를 제공합니다.

DSH 0.1.7-rc.2용으로 재구현한 TRAE 스타일 대화 롤백입니다. 턴별 파일 checkpoint로 변경 파일을 복원하고 같은 session id에서 모델 컨텍스트를 즉시 잘라냅니다. /rollback doctor가 framework-contract를 자체 점검하며 영향 파일 diff 미리보기는 재시작 후에도 정확하도록 현재 디스크 상태와 비교합니다. Ctrl+Shift+Z로 마지막 작업을 한 번에 되돌리고 롤백된 메시지 숨김 설정과 영어 UI 문구를 제공합니다.

작성기 통계 바닥글의 폭 제한을 해제합니다: 긴 턴 및 토큰 통계 줄이 말줄임표로 잘리는 대신 작성기 전체 너비에 걸쳐 표시됩니다.

Keeps a turn from closing on an unfinished reply: a stream guard turns silent truncation into an in-step retry, and a completion protocol asks the model to declare how each turn ended.

Adds a relay_check tool that runs relaycheck against an LLM API relay to detect model swaps, misreported usage, truncated context and injected prompts; keys are read from environment variables.

Idle-time context truncation for a single DeepSeek Harness session. After a configurable number of idle hours the older conversation is dropped without a summary, so one long-lived session cannot leak one task into the next; it ships a per-session toggle, a tab next to the chat, and a /truncate command that IM bridges can drive.

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.

DSH 세션 제목을 자동 지정합니다. 첫 prompt와 첫 assistant 답변을 바탕으로 “project｜task” 형식으로 이름을 붙여 prompt 잘라내기를 대체합니다. bundle patch로 기본 세션 제목 provider를 바꾸고 세션 행 메뉴에 미리보기 dialog가 있는 “AI rename”을 추가합니다. 설정에서 provider/model, 명명 규칙, 진단 event log를 관리합니다. unit test 49개.

DeepSeek Harness 에이전트의 MCP 서버, 스킬, 도구를 관리합니다. 실제 컨텍스트 상태(로드됨/잘림/제거됨)와 각 기능의 세션별 전환을 제공합니다.

각 skill이 모델 컨텍스트에서 차지하는 비용을 보여 주고 삭제 없이 catalog에서 숨깁니다. 디스크의 모든 skill에 대해 모델이 받는 정확한 catalog 행(description 정규화 및 길이 한도에서 잘림), 본문 크기, token 추정치를 나열합니다. agent가 받는 catalog message를 바이트 단위로 미리 보고 현재 session에 이미 게시된 사본과 비교하며, 잘못된 frontmatter 또는 우선순위 높은 root의 shadowing 때문에 catalog에 도달하지 않는 skill을 보고합니다. 숨기기는 frontmatter에 disable-model-invocation을 기록하지만 파일을 보존하므로 `/name`으로 계속 쓸 수 있습니다. ClawHub, Skills.sh, GitHub URL에서 가져오기 및 사용자 소유 skill 제거도 지원합니다.

사용자가 직접 게시한 QQ Zone 글을 로컬 라이브러리에 보관해 생각·꿈·리뷰(책/영화 추천) 세 열로 자동 분류한다. 각 열은 따로 수동 수정할 수 있고 한 열의 실행 취소가 다른 열에 영향을 주지 않는다. 비밀번호 잠금 가능한 비공개 웹 뷰어, upstream API가 약 500자에서 자르는 긴 글 복구, 연/월별 단어 수 집계, 오프라인 열람용 첨부 이미지 백업을 제공한다. 쿠키는 이 컴퓨터에만 저장된다.