Plugins

Plugin DSH Web de gestion des sessions : archivage/désarchivage, suppression physique garantie (fichiers transcript et balayage de ~/storages/*.json), réaffectation de workspace par glisser-déposer et édition des messages par branche (edit/retry/reroll avec historique de versions réversible).

Couche de validation d’expérience pour DSH : collecte sans jeton des signaux d’échec d’outil/nouvelle tentative/recherche, affinage par IA en livres de compétences vérifiés (vérification à trois niveaux), consultés au début de la tâche. Fournit 10 compétences d’essai et des données de référence (réussite des tâches complexes 100 % contre 60 % sans).

Routage multimodal transparent pour les modèles texte uniquement : chaque image de chaque appel de modèle est entièrement transcrite (OCR verbatim, données, zones d’incertitude, durci contre l’injection) par votre propre compréhenseur multimodal, avec réexamen ciblé via vision_relook et nouvelle tentative automatique en cas d’échec lié aux images. Aucun point de terminaison intégré, aucun identifiant emprunté.

Relance automatiquement les requêtes API échouées pour que les conversations DSH se poursuivent. Prend en charge des codes d'état et des codes d'erreur configurables.

Paquet de garde contre le mojibake pour DSH : détection de texte corrompu par caractéristiques avant écriture avec revue puis libération (une nouvelle tentative identique passe), plus audit d'encodage pwsh avant exécution (ANSI/GBK/UTF-16/Default). Windows et chinois d'abord.

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.

Read-only /btw side questions for DeepSeek Harness: forks the current session, injects the running turn's task, recent tool calls and partial answer so a mid-turn question has context, denies every tool call at execution time, and keeps a per-session topic history with follow-up, stop, retry and delete.

Sends desktop notifications when a DSH turn finishes or fails, and sends 继续 automatically after the model-retry budget is exhausted.

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.

Prépare les fichiers avant leur ouverture par read, read_image et les outils vision. Réduit les images de plus de 2048 px ou 8 MB (un nouvel essai à 1280 px dans le même tour), convertit BMP/TIFF en PNG et détecte les extensions erronées d'après les octets. Convertit docx/xlsx/pptx, PDF avec couche texte, csv/tsv et html en Markdown sans programme externe.

Lorsque la limite de contexte est dépassée et que la compaction intégrée abandonne, une nouvelle compaction est lancée automatiquement, puis le traitement est réessayé. En cas de nouvel échec, la session d’origine est préparée dans un package handoff pour continuer dans une nouvelle. Aucun changement de moteur ni UI ; la reprise de session doit être déclenchée explicitement par défaut, et l’originale reste en lecture seule.

Protection contre les crashs dans le processus : reprend les handlers d’erreur fatale qui feraient quitter DSH, identifie le plugin responsable dans la stack et désactive uniquement sa ligne dans la couche home patch, sans toucher au profile ni à sa configuration provider. La page d’erreur propose réessayer, garder désactivé ou désinstaller ; si aucun responsable n’est identifiable, ouvre une session safe-mode temporaire (lancements CLI et Web uniquement).

Refuse les attentes au premier plan des agents DSH : commandes d’attente (sleep, wait, tail -f, watch, ping sans -c, boucles infinies), outils bloquants (wait_agent, job_output avec wait: true), répétition rapprochée d’une même commande de contrôle et commandes exigeant une intervention humaine (sudo sans -n, ssh sans BatchMode, connexions, éditeurs interactifs, docker -it). Propose une reprise concrète en arrière-plan ou la fin du tour. Budget et fenêtre de sondage configurables ; langue du message selon system (zh/en). Aucune dépendance à l’exécution.

Relance automatiquement les requêtes modèle échouées : intercepte chaque erreur et rejoue la même étape après un délai de 2/5/10/20 s, afin que les incidents de passerelle et limites de débit ne terminent plus le tour.

Anneau de quota GLM Coding Plan dans le dock du composer : fenêtres de crédits 5 h et hebdomadaire, niveau, actualisation auto et nouvelle tentative en un clic. Nécessite l’abonnement et la clé GLM Coding Plan, ainsi que les identifiants d’organisation et de projet BigModel (l’API de quota d’équipe renvoie vide sans eux).

Ajoute l’action native « Restart DeepSeek Harness » au menu de la zone de notification Windows et au menu du compte de l’application DSH. Au démarrage, le plugin prépare un resources/app.asar modifié ; un assistant détaché sauvegarde l’original et le remplace après une fermeture normale. Le patch est réappliqué après chaque mise à jour de l’application. L’une ou l’autre action relance proprement l’application au lancement suivant. Windows Desktop uniquement ; les autres hôtes renvoient 409 not_desktop.

Bouton près du composer réécrivant le brouillon en prompt plus exécutable. Quatre presets, de la retouche légère à l’expansion forte, plus presets personnalisés. Change preset, modèle de prompt et effort de raisonnement depuis le bouton sans quitter le composer. Visible seulement si le brouillon contient du texte ; minuterie pendant l’amélioration et retry en cas d’échec.

Quota réel de l’abonnement ChatGPT/Codex dans DSH. Lit `/wham/usage` : fenêtres 5 h/hebdo, compte à rebours du reset, solde d’usage supplémentaire, groupes de limites et limite code review. Liste/utilise les crédits de reset émis par le serveur avec idempotency key stable, sûre aux retries ; plusieurs comptes. `/codex/models` donne le catalogue réel, vide s’il est indisponible (aucune liste intégrée). Connexion en un clic depuis Settings (autorisation navigateur ou device code ; collage du callback en secours) ; identifiants dans le credential service DSH. Les appels utilisent la route Codex intégrée `llm-pi-ai` sans enregistrer ni occuper de route provider. Si un autre plugin la possède, seule la console quota reste. Pastille de quota restant près du composer.

Plugin de réparation des ACL du sandbox Windows pour DeepSeek Harness : corrige l’échec grantWrite (SetNamedSecurityInfoW Win32 5) et relance la commande discrètement.

Transfère les historiques de chat WeChat vers des sessions DSH : une cible de partage Windows récupère le ZIP de transferts groupés ; une surcouche graphique permet de choisir espace et session. Le ZIP arrive comme message utilisateur avec pièce jointe que l’agent décompresse et lit. Normalise les extensions à partir des octets magiques (WeChat peut nommer .jpg des données PNG/HEIC) et conserve le lot en cas d’échec pour réessayer. Nécessite Windows 10/11, Developer Mode activé et l’installateur du dépôt pour enregistrer la cible.

Bridge IM unifié DSH : 15 adapters/12 plateformes. Adapte les hosts 0.1.2-rc.1–0.2.1-rc.2 par capability probing (enum 12 tags, sans branche de version) ; ffmpeg requis seulement pour la voix Feishu. Approbations/questions dans le chat d’origine : deny-by-default, initiateur seul, expiration 5 min. Image/fichier selon metadata modèle host ; fichiers renvoyés, retry borné (envoi incertain jamais répété). Tabs Tasks/Channels cèdent le slot à un autre plugin. Preset im-concise ; retire GenUI fences/raw HTML des chats texte, svg→PNG. Allowlist modèle. /relay envoie la réponse finale par account ID/alias, sans deviner l’ambiguïté. Secrets vers vault à sens unique, jamais renvoyés en réponse. npm: dsh-im-union.

Relie DeepSeek Harness au service mémoire Hindsight selon les conventions omp : banque partagée, balises de projet automatiques, rappel projet et global, outils pour modèles mentaux, doublons et relance des opérations de fond échouées. Serveur Hindsight requis (http://127.0.0.1:8888 par défaut).

Surveille les échecs consécutifs identiques d’un outil et peut suspendre le flux avant l’étape suivante du modèle. Préserve les résultats d’origine et les répétitions réussies, avec un rapport dans le chat de DSH Web.

Récupère les tours DSH interrompus : élargit le budget de nouvelles tentatives propre au produit et affiche l’attente par compte à rebours dans la même étape. Garde une sortie tronquée dans le même tour et peut ouvrir une reprise facultative après sa fin ; si le compositeur est vide, le bouton d’envoi lance cette reprise.