Plugins

Turns a stated need into a plugin choice: an agent skill that queries a catalog of the whole `dsh-plugin` topic (2600+ repos, recrawled every 6 hours, Chinese descriptions from Workers AI) through a ranked search API, then hands back the matching `dsh plugin add` command.

Keeps what the agent builds at runtime: turns a `cordis_define` dynamic package into a real installed plugin that survives restart, writing the bundle and registering the profile layer with no pnpm, no network, and no build authorization.

Configurable per-turn budgets for supported file mutations, limiting distinct files, mutation calls, and UTF-8 payload bytes before tool execution.

Metadata-first runtime profiler for DSH agents: turn/step traces, model and tool wall-time timelines, provider usage, retries, errors, and live or historical sessions.

Real-time stage, ETA & cost dashboard for headless dsh tasks: turns the black-box session event stream into a live multi-task grid (cordis plugin + standalone local dashboard, no API calls).

OS-level system notifications for the dsh web GUI: popups when a turn finishes or the agent needs your approval, answer, or plan review while the page is in the background.

Conversation-completion notifications for DeepSeek Harness: Windows toast + sound on turn done / error / goal, plus ask & approval alerts - background-only, with a volume-adjustable bell in the conversation header.

`describe_image` tool: a vision bridge that sends images to mimo-v2.5 through the opencode Zen API (credential `OPENCODE_GO_API_KEY`, free route first with paid fallback) and returns text descriptions for text-only models, with native passthrough and ImageMagick transcoding of SVG/TIFF/HEIC formats.

Auto-discovery vision bridge for text-only DeepSeek Harness agents: automatically finds an image-capable model from your configured providers and returns picture descriptions as plain text via a vision tool.

Reuses DeepSeek web's built-in vision mode for text-only models: the deepseek_vision tool drives the local deepseek-vision-cli browser automation (manual login helper, deep-think enabled, auto-closes browser) and returns image descriptions as text.

Routes chat images to a fixed OpenAI-compatible vision model, returns factual observations to the selected main model, and reuses session-scoped observations across replay, compaction, and restarts.

Trending-topics collector aggregating Baidu / Toutiao / Douyin / Weibo hot lists in one call (Weibo optional cookie), returning titles with heat values for content creators — tool: `sxs_news_hot`.

Bi-temporal memory backed by a SodaMem daemon: recall is contributed to the prompt while each turn is assembled and every closed turn is ingested, with no tool call and no model call on the read path; each recalled fact cites the turn it came from and carries validity dates, so superseded facts stop being retrieved.

Brings your existing Claude / ChatGPT history into DSH: instant zero-token import of the memories.json inside a Claude export, local Claude Code transcripts with no export needed, and conversations distilled through the model you already configured in DSH. Every memory carries verbatim evidence checked against the source by code — a paraphrase is dropped and counted, and evidence found only in the assistant turns is always routed to review rather than stored.

Cross-session event-cue memory for DeepSeek Harness: auto-extracts task/conversation cues and user language habits from every turn into a persistent global cue bank, then re-injects remembered context on topic-switch detection (keyword or vector matching).

Minimal-compatible first turn for Web root sessions: starts with persistent bash and str_replace_editor, removes automatic workspace and skill context, then restores the selected preset after the first tool call or reply; includes a persistent composer toggle.

DeepSeek wallet balance at the sidebar foot: a credit-card icon with the amount tinted by remaining balance (green over ¥2, amber ¥0–2, red below), refreshed on mount, per turn-end, and on click; a signed delta floats up and fades on each balance change.

Per-turn and session-wide LLM cost metering: a dollar cost line under each message, a session cumulative total with hover breakdown, and an LLM+web price auto-refresh tool.

Folds each DSH Web turn's thinking, tool calls and context into bordered process boxes cut at every piece of model text; a collapsed box shows its latest two items, the stacked lines above its border open that one box, the border itself folds it back, and user dialogue, interactive questions and model text stay outside.

Left-edge turn-marks strip in chat: one bar per user message, click to jump to that message, hover for a preview, active bar turns white.

Turn-index sidebar: one entry per user turn, click to jump, scroll-spy highlighting.

Pin assistant replies from the Web action strip and recall them into the next model turn (`/pin` `/recall`, with optional wake).

Git-aware write guard plugin for DeepSeek Harness: blocks whole-file writes that would overwrite the user's uncommitted changes, and reports what happened to protected files at turn end

SearXNG-backed web search provider for the DeepSeek Harness web capability seam (ctx.web) — self-hosted metasearch, no API key, no model turn per search