Plugins

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.

Recherche les échéances de soumission aux conférences CCF avec les outils ccf_deadlines, ccf_venue et ccf_update, la commande /ccf et un panneau de bureau qui exporte la liste en calendrier. Les dates proviennent d'une synchronisation en direct (CCFDDL est la source principale, recoupée avec une seconde ; un écart d'au moins 36 heures est signalé comme conflit de sources). Aucune donnée n'est incluse ; les 30 prochains jours restent dans le system prompt.

Manuel de jugement sur la dynamique structurelle des chaussées : 59 fiches en 8 chapitres signalent les titres qui exagèrent le contenu, les formules qui échouent lorsqu’on les copie et les chiffres qu’il ne faut pas citer hors de leur contexte.

Affiche près du composer DSH le quota restant du canal Our Free Model (OFM) actif et liste tous les canaux connectés dans un panneau. Signale ceux épuisés ; changer de modèle change aussi le canal affiché. Requiert dsh-our-free-model avec au moins un canal connecté.

Reçus de livraison par session pour artefacts locaux : compare les hash actuels à une référence enregistrée, distingue les affirmations de l’agent de l’acceptation explicite de l’utilisateur et signale les fichiers modifiés ou indisponibles. Exporte un reçu Markdown via un outil, une slash command ou le panneau de conversation.

Outils adaptatifs à la version de DeepSeek Harness : lit la version active dans le gestionnaire de plugins et reproduit les règles de compatibilité d’app-boot (`engines.dsh` d’abord, puis peerDependencies `@deepseek-ai/dsh*`, comparaison incluant les prereleases, exceptions de version exacte). Évalue chaque plugin installé et signale statiquement les API renommées ou dépréciées entre v0.1.6 et v0.2.1. Fournit panneau de conversation, outil d’agent et routes HTTP state/scan/report. Lecture seule, hors ligne.

Audite la dépense de budget de contexte à chaque tour : détaille AGENTS.md, catalogue de skills, schémas d’outils et contexte d’exécution ; signale comme gaspillage la réinjection de contenu inchangé ; classe le volume de descriptions lu par le modèle à chaque tour avec plus de 60 skills et l’inflation de schéma par plugin. Déterministe, sans LLM. Quatre outils dont un balayage intersessions.

Compare chaque affirmation terminée avec la sortie d’outil lue pendant le tour. Seulement si la citation qui démontre une contradiction apparaît dans cette même sortie, guide une correction unique durant le tour. Signale les affirmations non étayées dans le journal et stderr. Vérifie gratuitement d’abord les termes absolus et nombres discordants ; sans outil, le contrôle n’atteint pas le modèle.

Audit des credentials avant publication : parcourt l’arborescence et repère API keys, tokens, clés privées, chaînes DB et mots de passe codés en dur avant commit/publication/livraison. Sans dépendance, entièrement hors ligne ; seuls les six premiers caractères des constats restent visibles.

Relecture des textes destinés aux personnes. Les outils write/edit exigent le flag human_facing ; le plugin vérifie en arrière-plan selon les règles PROSE.md par couches, prépare des annotations, les rapproche du fichier complet puis les fusionne via /humanize, /humanize-loop ou post-compaction pendant que la session principale continue. /humanize-review traite à la demande les fichiers suivis par git.

Enregistre un outil `valuation_service` qui lit un workpaper tree JSON 01/02/04 et renvoie un schema unifié : six groupes de ratios, ratios prévisionnels, champs d’entrée DCF avec marqueurs explicites d’absence, PS/PE comparables et IRR. Traitement déterministe et local ; WACC, croissance, capex, D&A ou fonds de roulement manquants restent vides plutôt que d’être devinés, et les écarts d’hypothèses sont signalés pour examen.

Enregistre un outil `context_extract` dans DeepSeek Harness. Compresse les longs documents bruts (rapports financiers, contrats, livres comptables) en Markdown structuré dense et JSON : mesures clés, tableaux, clauses numérotées, entités et alertes de recoupement, pour que l’étape suivante n’ait pas à relire l’original. Détecte et traite séparément états financiers, contrats juridiques et livres d’entreprise ; rapproche bilan et trésorerie et signale les écarts sans les corriger silencieusement. Lecture seule, sans modifier le fichier source. Nécessite Python avec openpyxl et python-docx.

Studio de skins utilisateur War Thunder : sélection de quadrilatère UV par quatre points, calques graphiques découpés selon le quadrilatère et export PNG/DDS/.blk.

Ajoute au compositeur un bouton ✨ et la commande /enhance pour structurer le brouillon, signaler les questions ouvertes sous le champ, suivre le modèle de la session et permettre l’annulation.

Démineur dans une fenêtre flottante déplaçable, avec première case sûre, drapeaux, ouverture en accord et chronomètre.

Mémo par semaine ISO : navigation semaine/mois/trimestre/année, archive trimestrielle, analyse IA et export Markdown. Modifier une semaine passée exige force flag ; analyse les pannes telemetry de la suite dans un rapport d’issue GitHub.

Signale les paires ICD à examiner (combinaisons dagger/asterisk et combination codes) selon GB/T 14396-2016 et les notes publiques. Il les relève sans trancher ; le codeur décide.

Reçus par session des artefacts locaux : compare les empreintes actuelles à la base enregistrée, distingue déclarations de l’agent et acceptation explicite, signale fichiers modifiés/indisponibles et exporte un reçu Markdown via outil, slash command ou panneau.

Extrait du texte le numéro et la date de facture, les parties, les identifiants fiscaux ainsi que le montant, la taxe et le total. Vérifie que montant plus taxe égale bien le total et signale les documents incohérents.

Vérificateur de compatibilité des plugins dsh : compare les peerDependencies des plugins installés à la version dsh active pour repérer ceux que le contrôle préalable au démarrage refusera et affiche des commandes de correction à copier. Une bannière en haut de dsh web permet d’isoler, désinstaller ou mettre à niveau ; sauvegarde automatique avant changement et retour arrière si la vérification échoue.

Badge du solde DeepSeek API en bas de la barre latérale, à côté de Settings : lit la clé déjà stockée par DSH, affiche au survol le détail entre crédit accordé et rechargé, avertit sous ¥2 et s’actualise toutes les 60 secondes ou au clic.

Épingle les sessions dans une unique section Pinned au-dessus de tous les groupes d’espaces de travail dans la barre latérale Web, sans toucher à leur ordre stocké ; l’épingle est un indicateur dans les réglages hôte. Le menu contextuel permet renommer, épingler, bifurquer, archiver, supprimer, copier le titre ou l’ID ; les libellés suivent la langue de l’interface.

Intégration Git pour DSH inspirée du VCS d’IDEA : l’icône Git près du mode de permissions affiche branche et changements en attente ; survol = sélecteur de branche, clic = panneau VCS façon IDEA (arbre, graphe de commits, changements/indexation, détails de commit, diff). Réglages d’identité de commit, binaire Git et indicateurs réseau. N’enregistre aucun outil modèle dans DSH.

QQ Mail et toute boîte IMAP/SMTP dans DeepSeek Harness : les outils Agent `qqmail_*` permettent de lister, rechercher, lire, envoyer, répondre, marquer et télécharger des messages via IMAP/SMTP avec un code d’autorisation QQ Mail (sans OAuth ni redirection 2FA). Inclut la CLI `qqmail`.