Plugins

Guarda de limpeza irreversível de segredos: reescreve chaves de acesso, tokens bearer e blocos de chave privada como marcadores `[REDACTED:<category>]` antes que cheguem ao log da sessão e ao modelo.

Gestor de contexto para o DeepSeek Harness: resume automaticamente cada turno em registos ordenados por prioridade, injeta-os juntamente com notas fixadas entre sessões e texto personalizado em cada pedido ao modelo, e reduz intervalos reais de conversa de qualquer tamanho a resumos para poupar tokens. Persistente e seguro face a reinícios. NOTA: instalação manual em vários passos — copie as três pastas do pacote para profiles/node_modules e adicione as linhas fornecidas no cordis.patch.yml, depois reinicie o DSH; não é uma instalação padrão com um único comando.

Invoca Skills acessíveis ao utilizador com tokens @name e mostra sugestões de Skills correspondentes no compositor da conversa.

Widget de saldo da API DeepSeek para a barra lateral web do dsh: uma pílula de saldo em tempo real e autoatualizada (saldo / gasto de hoje / tokens de hoje) com popover de detalhe (gasto acumulado, uso mensal, contagem de pedidos). Cada utilizador vê apenas o seu próprio saldo; as chaves são reso

30 paletas de temas clássicos de código aberto (Catppuccin, Nord, Tokyo Night, Gruvbox, Solarized etc.) mapeadas para os tokens oficiais do ThemeService do DeepSeek Harness, com um painel de Style Settings para ajustes ao vivo de cor/fonte/raio e exportação/importação JSON para compartilhar pacotes de configuração.

Catch-em-all estilo Pokemon: seu trabalho real é a grama alta — turnos, chamadas de ferramenta e erros geram encontros selvagens; jogue bolas, preencha uma dex de 28 slots, monte um time de seis. Zero tokens, invisível para o modelo.

Aviso amarelo de uma linha para um erro de cache de prompt após a reconstrução do cache; mostra tokens de entrada sem cache cobrados novamente, tokens atingidos no cache e TTFT.

Versão aprimorada do dsh-skin: oferece troca de skins e papel de parede de fundo personalizado para o DeepSeek Harness. Em relação à versão original, corrige dois problemas — ① upload de imagens grandes/formatos especiais sem resposta (agora usa createImageBitmap para decodificação direta, com indicadores de processamento e erro); ② ao trocar de tema, overrideTokens disparava theme/change causando estouro de pilha por recursão infinita (adicionado um guard de skin/claro-escuro para quebrar o ciclo). O armazenamento de papéis de parede passou do esquema comprimido em localStorage para arquivos originais salvos diretamente no IndexedDB, com exibição em qualidade total, zero compressão, e migração automática dos dados antigos

Side-by-side model comparison for the DeepSeek Harness Web UI: send one prompt to 2–4 models in sibling sessions forked from the current one, watch the answers stream in columns (tabs on phones) with latency, time to first token, tokens, and cost, then co

DSH Web plugin that injects an SVG (feTurbulence + feDisplacementMap) refraction filter and applies it through backdrop-filter to panels, bubbles and the input card, and bundles the orca-scene skin with theme tokens plus light and dark backgrounds.

Token usage dashboard in the DSH settings page: nine metric cards (total, today, yesterday, peak day, longest session, active-day streak, active days, sessions, subagent tokens), a GitHub-style daily/weekly/cumulative heatmap, and an activity insight list, aggregated locally from session logs and projection caches, keeping token history after a session is deleted.

Fronts stdio MCP servers with three router tools (list_capabilities, describe_tool, call_tool) so schemas load on demand instead of riding every request; shapes every tool result through a deterministic output ladder that archives the original first (retrieve_spill reads it back); prices tool groups in tokens and can hide built-in groups per agent. Bridges to a standard-library Python router, so Python 3.9+ is required.

Model-callable compact_agents tool that force-compacts context in every live session (main session, ordinary sub-agents and AgentTeams members alike), queues busy targets and compacts them when the turn ends, reports the shadowed node count and estimated tokens per target, and ships a browser form for the trigger ratio, retained ratio, controlled-phase output budget, auto-continue budget and an end-of-turn pre-compaction ratio that compacts near the trigger line once a turn ends, so a reply no longer opens with a wait for the summarizer.

Same-origin in mechanism with the two loudest lines in context compression. The code-skeleton gate follows the skeletonization approach of Headroom (Apache-2.0), whose published headline is 20% fewer tokens for coding agents and 60–95% fewer tokens for JSON, same answers. The estimator channel follows TokenPilot (arXiv:2606.17016), which reports up to 60% lower cost for long-session agents. Both figures are theirs, quoted as-is; this plugin ships no benchmark of its own and claims no reduction of its own. What it adds for DeepSeek Harness: choose a compression profile, set the Auto Compact trigger level and toggle code-skeleton compression from one settings section, with exact DeepSeek V4 tokenizer measurement, same-revision count verification, and fail-open behaviour that keeps the original tool results on unsupported models.

DeepSeek Harness Web 悬浮窗：实时显示会话输出吞吐（tok/s）、本步/累计 token 与首 token 延迟，支持实时/均速双口径与多会话切换。

Kaze Tachinu (The Wind Rises) skin for the DSH Web GUI: a full-screen film wallpaper behind glassmorphism surfaces, sky-blue interaction tokens, redrawn vector logos for the sidebar and home hero, a restyled composer card with themed placeholder text, and a bottom fade on the message list.

Out-of-tree dsh plugin: GitLab, GitHub, Gitee, Gitea, and Bitbucket tokens stay out of the model context, stored encrypted (AES-256-GCM) in a plugin-owned file; the model calls forge API tools on demand, and the web settings page manages sites and tokens.

Video-style turn navigation for the Web UI: full-map progress bar with hover preview cards (tools/files/tokens), drag scrub, ⌘K in-conversation search, ⌘↑/⌘↓ stepping, Esc-return. Event-log powered — works beyond the loaded window. Dual-half plugin on the official composer.dock slot.

Detects WSL2 clock skew versus Windows that can break TLS and tokens.

Gemini-styled skin restyling the conversation ground, sidebar rows, hero greeting and composer through one stylesheet over the --dsw-* tokens, with light and dark both covered.

Draggable ring gauge showing live model output speed (tok/s) with an expandable per-step detail panel (exact speed, TTFT, cache-hit rate, cache read/write, reasoning and input tokens, model, cumulative tokens).

Auto-continue for the dsh web GUI: when a local model hits its output-token cap and the max-tokens notice appears, the plugin automatically queues a short continue message so the reply keeps flowing — dual-source truncation detection, loop protection, and no UI.

DeepSeek account-wide usage & balance panel from the official platform API (api/v0/usage/by_api_key): cache hit/miss & output tokens, cost, 24h–90d ranges, persistent platform token, and a model-visible `deepseek_usage_query` tool.

Keeps the native per-message status line always visible without hovering — date and time on every message, run duration, TTFT, and tokens per second on assistant messages.