Plugins

Widget de coût des tokens en direct par tâche : fenêtres tarifaires pleines/creuses, coûts projetés après hausse, tableau des coûts par tâche triable, multidevise (COP/USD/CNY + ~40 autres) avec taux de change, et explication LLM en streaming en un clic des tâches coûteuses.

Tableau de bord d'usage par fournisseur : indicateurs de jetons, de cache et de sortie par fournisseur, carte de chaleur de 53 semaines, graphique de tendance avec vue horaire « aujourd'hui », exploration par modèle, estimation des coûts, export CSV et cartes de santé (TTFT, vitesse de génération, contexte de pointe, taux d'erreur).

Plugin fournisseur Command Code pour DeepSeek Harness (dsh) — fonctionne avec tous les plans Command Code, y compris le plan Go à 1 $ ; utilise la clé API d'authentification Studio via /alpha/generate, pas la Provider API.

Bundle de plugins DeepSeek Harness : outils qwen_vision (compréhension d'image Qwen-VL) et qwen_generate (texte-image Qwen-Image) pour les modèles texte seul.

Solde et usage de session en direct pour l'interface web DSH : bouton de solde dans l'en-tête (à gauche du journal de session), menu déroulant avec le solde du compte, ventilation des jetons en temps réel (échec/succès de cache, écriture de cache, sortie), taux de succès du cache et estimations de coût aux tarifs officiels avec bascule automatique heures pleines/heures creuses.

Un plugin DeepSeek Harness intégrant claude-mem : requêtes, injection de contexte, sauvegarde manuelle et cycle de vie de session via l'API HTTP d'un worker claude-mem local.

Read-only agent checkup with four probes (output repetition, error visibility, tool failure rate, context bloat) over recent local session logs, rendered as an HTML report.

Long-term memory kept as plain Markdown plus git: each session opens with a generated memory-context block assembled from a global store and per-project books (role card, user profile, current focus, routing table, recent logs), the plugin records which memory files the agent read and reminds it to extract conclusions while the day's log is unextracted, personas live in roles/<role>/SOUL.md role cards, and it ships a seven-tab settings page in the sidebar plus four skills. Requires Python 3.10+.

Configurable personas, separate reply generation, and evidence-backed memory with review and correction for DeepSeek Harness.

Model rotation for DeepSeek Harness: groups each provider's models by source, orders them by hand or by measured success rate and latency, stores model plus reasoning-effort presets, falls over to the next entry when a model hits its quota or is unavailable, resends the same step on the next model, and shows which model actually answered. Uses the models of whatever provider plugins are already installed (Antigravity accounts, Codex subscription, DeepSeek API keys); it does not provide model access itself.

Usage statistics settings page computed locally from session logs: 30-day daily list, per-model and per-session tables with provider-aware pricing (DeepSeek peak/off-peak, bundled TokenBom flat rates); skips inherited events of forked subagent sessions to avoid double-counting.

Title-bar balance badge for DeepSeek Harness: account balance, today's cost, session tokens, and peak/off-peak pricing awareness (Beijing-time weekday windows) with flat-rate support for custom providers like TokenBom.

Deepseek harness plugin for control rate limits

An animated macOS desktop companion for DeepSeek Harness that follows work across chats, reacts to approvals and errors, takes snack breaks during long runs and naps when idle. Requires a separate shell app.

A draggable basketball desktop pet that bounces around the window, with a hoop to score in and a crate that keeps the ball.

OpenSpec spec-driven workflows for DeepSeek Harness, generated from OpenSpec 1.14.3: twelve /opsx-* slash commands that carry the workflow prompt plus the arguments you type, and twelve openspec-* skills the model loads on its own.

Mailbox for DeepSeek Harness: sessions pass letters through a shared folder, and an hourly randomized scheduler can wake a session on its own (mutual letters between two sessions, or a gentle nudge to the owner). Per-recipient hourly rate limit and a UTF-8 log.

Image generation through TokenRhythm: mounts a no-npm-dependency Python stdio MCP server as mcp__tokenrhythm_image__generate_image, producing images with qwen-image-2.0 or wan2.7-image and saving them to a directory you choose. Needs Python 3.8+ on PATH and TOKENRHYTHM_API_KEY.

Text-to-image through a StepFun-compatible endpoint: mounts a no-npm-dependency Python MCP server behind an NDJSON bridge as mcp__simple_image__simple_image and returns the generated image URLs. Needs Python 3.8+ on PATH and STEP_API_KEY.

Configurable, provider-agnostic image generation: mounts the bundled MCP server behind an NDJSON bridge as mcp__image_generation__image_generate and mcp__image_generation__image_config_status, and registers the bundled image-generation skill. Needs baseUrl and apiKey; without them the MCP server mounts nothing.

Brings McDonald's China official MCP (mcp.mcd.cn) into DSH. Paste a token you generated yourself and the plugin handshakes, calls tools/list, and registers every remote tool as a native DSH tool (mcd_ prefix) — upstream additions and removals need no code change, just a restart. Ships four always-on tools (mcd_status, mcd_test, mcd_tools, mcd_call) for self-checks and direct fallback calls, plus a settings page with a one-click connection test. The token is stored only on this machine in ~/.dsh/dsh-mcd-mcp.json (mode 600); it never enters the repository or the conversation context.

One Settings page for every DeepSeek Harness session: browse active and archived sessions by workspace, generate a title from the whole conversation for live or archived sessions, archive or restore, move a cold session to another workspace, and permanently delete an archived session (with its subagent descendants after confirmation).

Connects CANNBOT gateway models to DSH by reusing the sign-in CANNWork already establishes for opencode: the JWT and virtual key are read from disk on every request, so no separate OAuth flow or extra key is needed, and the account's visible models are registered in the model picker.

Writes the session's estimated cost into the composer token-usage pill, and adds a per-bucket amount plus an estimated total to the usage dialogs, using DeepSeek's official peak and off-peak rates in yuan.