Plugin

Plugin DSH Web per gestire le sessioni: archiviazione/ripristino, eliminazione fisica garantita (file transcript e scansione di ~/storages/*.json), riassegnazione del workspace con trascinamento e modifica dei messaggi basata su branch (edit/retry/reroll con cronologia delle versioni reversibile).

Livello di validazione dell'esperienza per DSH: raccolta a zero token di segnali di errore degli strumenti/retry/ricerca, raffinamento tramite AI in manuali di competenze verificati (verifica a tre livelli), consultati all'avvio del task. Include 10 competenze di prova e dati di benchmark (successo su task complessi 100% contro 60% senza).

Routing multimodale trasparente per modelli solo testuali: ogni immagine in ogni chiamata al modello viene trascritta completamente (OCR letterale, dati, zone di incertezza, protetta dalle iniezioni) dal tuo stesso comprenditore multimodale, con riosservazione mirata tramite vision_relook e nuovo tentativo automatico in caso di errori legati alle immagini. Nessun endpoint incluso, nessun login preso in prestito.

Riprova automaticamente le richieste API non riuscite per non interrompere le conversazioni DSH. Supporta codici di stato e codici di errore personalizzabili.

Bundle di protezione dal mojibake per DSH: rilevamento del testo corrotto basato su caratteristiche prima della scrittura con revisione-poi-rilascio (un nuovo tentativo identico passa), più audit della codifica pwsh prima dell'esecuzione (ANSI/GBK/UTF-16/Default). Windows e cinese prima di tutto.

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.

Prepara i file prima che read, read_image e gli strumenti vision li aprano. Ridimensiona immagini oltre 2048 px o 8 MB (un nuovo tentativo a 1280 px nello stesso turno), converte BMP/TIFF in PNG e riconosce upload etichettati male esaminandone i byte. Converte docx/xlsx/pptx, PDF con livello di testo, csv/tsv e html in Markdown senza programmi esterni.

Quando si supera il limite di contesto e la compaction integrata si arrende, esegue automaticamente un’altra compaction e poi riprova. Se fallisce ancora, organizza la sessione originale in un pacchetto handoff per continuare in una nuova. Nessun cambio engine o UI; la ripresa della sessione richiede un’attivazione esplicita per impostazione predefinita e l’originale resta in sola lettura.

Protezione crash in-process: prende in carico gli handler di errore fatale che terminerebbero DSH, identifica il plugin responsabile dallo stack e disattiva solo la relativa riga nel livello home patch, senza toccare profile o configurazione provider contenuta. La pagina di crash offre riprova, mantieni disattivato o disinstalla; se il colpevole non è identificabile apre una sessione safe-mode temporanea (solo avvii CLI e Web).

Rifiuta le attese in primo piano degli agenti DSH: comandi di attesa (sleep, wait, tail -f, watch, ping senza -c, cicli infiniti), strumenti bloccanti (wait_agent, job_output con wait: true), ripetizione dello stesso comando di controllo entro una finestra e comandi che richiedono input/conferma umana (sudo senza -n, ssh senza BatchMode, login, editor interattivi, docker -it). Suggerisce un’alternativa concreta: eseguire in background o terminare il turno. Budget e finestra di polling configurabili; lingua secondo system (zh/en). Nessuna dipendenza runtime.

Riprova automaticamente le richieste modello fallite: intercetta ogni errore e ripete lo stesso passaggio dopo attese di 2/5/10/20 s, evitando che brevi guasti gateway e limiti di frequenza terminino il turno.

Anello della quota GLM Coding Plan nel dock composer: finestre credito di 5 ore e settimanale, livello, aggiornamento automatico e nuovo tentativo con un clic. Richiede abbonamento e chiave GLM Coding Plan, oltre agli ID organizzazione e progetto BigModel (senza, l’API quota team restituisce vuoto).

Aggiunge la voce nativa “Restart DeepSeek Harness” al menu della tray di Windows e al menu account dell’app desktop DSH. All’avvio il plugin prepara un resources/app.asar modificato; un helper separato salva l’originale e lo sostituisce dopo una chiusura normale. La patch viene riapplicata a ogni aggiornamento desktop. Entrambe le voci riavviano correttamente l’app al successivo avvio. Solo Windows desktop; gli altri host restituiscono 409 not_desktop.

Pulsante accanto al composer che riscrive la bozza corrente come prompt più eseguibile. Quattro preset da ritocco leggero a espansione aggressiva, più preset personalizzati. Cambia preset, modello prompt e sforzo di ragionamento dal pulsante senza uscire dal composer. Visibile solo con bozza non vuota; mostra timer e retry se fallisce.

Mostra in DSH la quota reale dell’abbonamento ChatGPT/Codex. Legge `/wham/usage`: finestre di 5 ore e settimanali, conto alla rovescia reset, saldo uso extra, gruppi di limiti aggiuntivi e limite code review. Elenca/riscatta crediti di reset server con idempotency key stabile e sicura ai retry; supporta più account. Legge `/codex/models` per il catalogo reale e mostra vuoto se indisponibile, senza lista integrata. Accesso one-click da Settings (autorizzazione browser o device code, fallback con incolla callback); credenziali nel credential service DSH. Le chiamate riusano la route Codex integrata `llm-pi-ai` senza registrare né occupare una route provider. Se la route appartiene a un altro plugin, resta solo la console quota. Chip quota residua accanto al composer.

Plugin di riparazione ACL del sandbox Windows per DeepSeek Harness: corregge l’errore grantWrite (SetNamedSecurityInfoW Win32 5) e riprova il comando senza mostrarlo.

Inoltra le chat WeChat alle sessioni DSH: una destinazione di condivisione Windows acquisisce lo ZIP dei messaggi inoltrati; un overlay GUI consente di scegliere workspace e sessione. Lo ZIP arriva come messaggio utente con allegato che l’agent estrae e legge. Normalizza le estensioni dai magic byte (WeChat può chiamare .jpg dati PNG/HEIC) e conserva il lotto se l’invio fallisce, per riprovare. Richiede Windows 10/11, Developer Mode attivo e installer del repository per registrare la destinazione.

Bridge IM DSH: 15 adapter/12 piattaforme. Adatta settings host 0.1.2-rc.1–0.2.1-rc.2 solo via capability probing (enum 12 tag, nessun ramo per versione); ffmpeg di sistema serve solo al tier vocale Feishu. Pannello 25 comandi. Approval/domande si rispondono nella chat d’origine: deny-by-default, solo initiator, scadenza 5 min. Input immagine/file secondo metadata del modello host; file restituiti, retry limitato (invio incerto mai ripetuto). Tasks/Channels cedono lo slot già occupato da un altro plugin. Preset im-concise, rimuove GenUI fence/raw HTML dalle chat solo testo, svg→PNG. Allowlist modelli. /relay invia risposta finale al notify channel via account ID/alias, senza indovinare ambiguità. Secrets solo in uscita verso vault, mai restituiti nelle risposte. npm: dsh-im-union.

Collega DeepSeek Harness al servizio memoria Hindsight secondo omp: banca condivisa, tag progetto automatici, richiamo di progetto e globale, strumenti per modelli mentali, duplicati e retry di operazioni in background fallite. Richiede server attivo (default http://127.0.0.1:8888).

Rileva errori consecutivi identici degli strumenti e può sospendere l’esecuzione prima del passaggio successivo del modello. Conserva i risultati originali e le ripetizioni riuscite, con un report nella chat DSH Web.

Recupera turni DSH interrotti: amplia il budget di retry del prodotto e mostra l’attesa con conto alla rovescia live nello stesso passaggio. Mantiene un output troncato nello stesso turno e può aprire una continuazione facoltativa dopo la fine; con composer vuoto, il pulsante invio avvia la continuazione.