Plugins

Analisa imagens de conversa por meio de prompts específicos por modo (caption, UI, documento, grounding, topologia etc.) e injeta evidências estruturadas com primitivas de coordenadas (caixas, pontos, referências) como texto, com cache em nível de sessão para reutilização em replay e compaction.

Permite que um agente local do DeepSeek Harness use o BailingHub para consultar dados, atualizar registros e executar ações permitidas em uma loja online, SaaS ou outro sistema administrativo.

Árvore de arquivos na barra lateral para DeepSeek Harness: um painel de arquivos com uma árvore preguiçosa com ícones por extensão, ações de copiar caminho e enviar para o agente.

Mascote baleia cibernética para DeepSeek Harness: uma baleia flutuante, alimentável e personalizável em cores sobre shell.overlay que relata uso de tokens, cota, interações e contagens de sessão — um único bundle dsh-plugin carregando o cérebro de chat do host e a superfície do navegador.

Ícones em estilo Fluent automáticos e personalizáveis pelo utilizador para Settings e ações da barra lateral do DSH, com pacotes SVG offline e preservação dos ícones originais.

Central de habilidades na interface web do DSH: navegue pelas habilidades carregadas agrupadas por origem (system / project / user / custom / runtime), com criação, ativação/desativação e exclusão (move para .trash).

Botão de iniciar/parar num clique fixado no canto inferior direito da janela: reinicia o DeepSeek Harness no local através de um auxiliar desanexado que relança a invocação de arranque exata (protegido por mesma origem em loopback), com uma ação opcional apenas para parar.

Consultor de avaliação e melhoria de PR: avalia a probabilidade de sucesso de PR com base em antipadrões históricos e padrões de sucesso, com pontuação num clique e sugestões acionáveis.

Kit de visão para DeepSeek apenas texto: ferramenta `vision` invocável pelo modelo, adaptadores wrapper para deepseek/opencode-go (v4 flash/pro), cota do Antigravity IDE (default, flash/pro) / qualquer VLM compatível com OpenAI / Gemini / canais locais do Ollama, memória de evidências com reidratação por compactação, cache de hash de conteúdo e um painel de cliente bilíngue.

Testes de aceitação independentes no navegador para entregáveis de agentes: entra uma especificação JSON e sai um veredito em Chromium real.

Redige padrões de segredos suportados na cópia exportada de `session-telemetry/record` antes que os backends de telemetria configurados a recebam.

Personal dsh preferences: a tunable companion to the official compaction engine (threshold, retention, summarization model/reasoning, turn-end compaction), llama.cpp model import for the Models page, and tab/notification helpers (status-light favicon, bro

DSH plugin: counts completed compactions per session and raises a floating reminder — with bilingual UI — only after repeated compactions plus heavy extra token spend.

Windowed, reversible context management for DeepSeek Harness; replaces Basic compaction.

Deterministic project actions shared by people and agents: one actions.json, a right-sidebar run panel, and agent tools.

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.

Compresses older DSH context into a structured handoff, retains recent messages, and keeps session history searchable.

Animated DSH Web pet with model-adaptive output pacing, tool-state reactions, replaceable character packs, and a bundled skill for creating or extending pet animations.

Cross-session experience memory for DSH. Lessons are evidence-graded, injected when the turn is about the same thing, and shown just before an action only when the record declares which call it applies to. A rule that holds whatever the turn is about — "always answer in Chinese" — can be marked **standing** and is then carried every turn: measured on 24 isolated real-model turns, a standing rule is followed 6/6 against 0/18 for the controls (empty store, same record unmarked, unrelated standing record). A guess still cannot become always-on: a standing record needs a verified passage like any other. What nobody looks up is retired, reversibly. Five tools, seven slash commands, zero runtime dependencies.

Simulated life context for DeepSeek Harness: every turn injects the past 24h of simulated life events from the workspace .life/ directory (sliding-window filter, session-level deduplication with delta updates, world summary) as a collapsed context notice, and the life_react tool lets the agent record feelings, thoughts and actions back into its life log for the next round of world evolution — turning a DSH agent into a roleplay character with a persistent daily life.

Compaction fallback for compaction-basic: when its LLM summary throws, comes back empty or cannot fit the window, the span is compiled deterministically with a seq pointer on every line (vendored dsh-compaction-instant compiler). Adds recall and search tools, /recall and /context commands, an idle trigger at 45% of the window and an on-disk span archive. The engine mounts as one row in the agent preset (docs/dsh-integration.md).

Model-autonomous context compaction for DeepSeek Harness: a mind_map tool surveys the agent context surface (stable seq ids, roles, token weights, clear-boundary markers, the clear-mind playbook), a clear_mind tool replaces a chosen conversation-history span with a model-written checkpoint through the native compaction transaction, call and result pairs self-collapse into one-line tombstones, and a proactive reminder nudges the model when the context grows long or a turn runs too many steps.

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.

Distills session work into background observations and reflections to preserve context across compaction.