Plugins

Drive the user's real browser through the local Kimi WebBridge daemon: navigate, find-tab, snapshot, click, fill, evaluate, CDP, screenshot, network, upload, PDF export and tab-management tools that reuse existing login sessions.

Playwright MCP browser control for AI agents: drive a real, visible Chrome (navigate/click/type/snapshot + ~80 browser_* tools), with a live screenshot panel embedded in the GUI that refreshes after every action.

Lets sandboxed agents reach the web through your local proxy: net_fetch fetches pages, net_proxy_status reports proxy status.

Route eligible shell commands through rtk to compress tool output and save tokens.

Pause / resume / cancel the in-flight conversation task from the composer area: forced pause interrupts immediately and remembers the interrupted tool, safe pause lands after the tool or reasoning finishes, resume asks for confirmation, and the default pause granularity is configurable in the settings.

A GrokBot-style animated avatar with a synchronized typewriter status line — right in front of the "Deep diving..." indicator.

Arknights-inspired fan skin for the DSH Web UI featuring Pramanix and Eyjafjalla in a day/night celestial garden.

Theme studio for the DSH Web GUI: five built-in presets plus fully customizable light/dark palettes (accent, background, foreground, UI and code fonts, translucent sidebar, contrast), hot-swapped instantly and persisted in localStorage.

Neo-brutalism skin with two switchable schemes (Blue Command / Aged Newspaper), hard shadows, sharp corners, and light/dark theme support.

OpenCode Go balance pill at the right end of the composer tool row: rolling/weekly/monthly quota remaining straight from the Zen usage endpoint, with hover detail and warn/danger coloring.

Codex-style auto-hiding right-edge turn rail: edge-peek show/hide, per-turn tooltip, in-rail search, and a scrolling 30-turn window for long sessions.

Automatically folds tool calls into Think-segment groups and collapses each finished turn into a single header showing duration, token, throughput, and cache-hit metrics.

Left-edge user-Turn navigation rail with reading-position highlighting, hover previews of per-turn metrics and model-answer excerpts, keyboard and click-to-jump controls, and local conversation search.

Right-hand widget rail for the DSH Web UI: live session stats (turns, LLM/tool time, TTFT, speed, cache, tokens) and OpenCode Go quota via a same-origin host proxy; extensible widget registry for more platform usage widgets, visuals, utility actions and external integrations.

DSH Web right-side output dock that collects session artifacts, previews user-facing files, and preserves tab state per session.

Menu de session au clic droit et gestionnaire de sessions dans la barre latérale : épingler, renommer, archiver, forker, exporter, copier le cwd/id/lien profond, et ouvrir dans l'explorateur ou une nouvelle fenêtre.

Plugin DeepSeek Harness (DSH) : un sélecteur de « langue de réflexion » dans Réglages → Général qui indique au modèle dans quelle langue raisonner (chaîne de pensée) via une section du prompt système. Aucun appel supplémentaire, aucune latence, 12 langues.

Adaptateur de fournisseur Command Code Go pour dsh : scanne dynamiquement le catalogue de modèles du plan Go et sert /alpha/generate via la jonction LLM du harness.

Plugin fournisseur optionnel pour dsh-skill-pack-security : enregistre le dossier skills/ (zh) ou skills-en/ (en) du pack dans ctx.skills. Les deux éditions sont embarquées dans pack/.

Panneau développeur latéral droit pour DeepSeek Harness : fichiers, historique de session et contrôle de version git.

Analyse d'usage légère pour DeepSeek Harness — tendances de tokens, cartes de chaleur d'activité, répartition par modèle et exports.

Pre-flight plugin-library check for DSH agents: task-type requests trigger a Top-K hint of suitable plugins with use_when/not_for capability declarations, making plugin usage predictable instead of opportunistic.

Session repair button in the conversation header: detects & fixes session-log corruption — tool-call id swaps, empty call ids (`message must have tool source`), unknown event types from disabled plugins (marks `ignorable`), torn zstd tails, missing final-frame newlines; ghost-style UI, auto-backup before writes.

A 轨迹解读 (trajectory interpretation) tab that summarizes each user round — what was wanted, how the assistant thought and executed, files/commands/errors — via a rules engine plus optional LLM narrative; user messages stay verbatim.