Plugins

Every night generates the companion character's schedule for the next day with DSH's configured model, from the persona (AGENT.md by default) and a weighted day type (work, rest, outing, home, overtime, slacking, or your own); injects the date, time, current activity and what comes next into the system prompt, with a speaking-style note so the character does not recite a timetable; adds schedule_view and schedule_adjust tools and a /life command that turns the injection on or off per session. With aa1355721445-lgtm/dsh-image-studio installed, each entry gets an outfit, scene and props from its libraries. A settings page shows today's and tomorrow's timelines and edits day types.

A character who lives on her own phone: the plugin installs and runs her Android phone, and any app notification can wake her; she installs apps herself and browses, likes and comments in apps such as Xiaoheihe or Bilibili; she has her own QQ (NapCat / SnowLuma, multi-part replies, stickers); memory with people cards, a daily diary at 23:30 and automatic recall before each wake (local embeddings by default, SiliconFlow / DashScope optional); a life between conversations (after HDS-Interlude); and an optional microVM computer with approval gates. Sidebar tabs show her phone, her computer and a dot panel; personal settings live in ~/.dsh/dsh-dot.json, and each of the six components can be turned off. Only the QQ number set with /qq owner counts as you; other people's messages are treated as untrusted notifications.

Agentic answer reviewer for DeepSeek Harness: each final assistant turn is re-reviewed by a separate LLM graded 1-100; below-threshold scores steer the agent back with concrete feedback. The score is shown as a chip on the answer's action row (green pass / red needs-work, with retry count and reason on hover). Live gate config from the Reviewer 配置 tab in the conversation view ring above the composer (对话 / 轨迹 / 记忆系统 / Reviewer 配置), from a 127.0.0.1 HTTP server, or from an optional right-sidebar tab via dsh-better-sidebar.

Trellis-style workflow for DeepSeek Harness: per-turn phase breadcrumbs via agent/pre-step, bundled skill provider, and DSH-native task/phase state with git-cleanliness guardrails.

Fetches any http/https page and returns title, description, main text, heading outline and links, with stats (char/word counts, reading time, language) and top keywords; can download content images to the session workspace for vision analysis and dispatch them to registered analyzers via the scrape.imageAnalyzer service.

Fork of dsh-superagent for local use: each sidebar workspace gets a main session that turns a request into a ticket and has a worker session do the work inside that workspace's directory. Cannot be installed alongside upstream dsh-superagent: both use the package name dsh-superagent and mount the same plugin ids.

Plays Inception Mercury's diffusion (denoising) stream live in the dsh Web UI answer bubble: for routes that support it, the request is re-sent with `diffusing: true` and the real frames are replayed from noise to final text, while only the converged answer reaches the session log. Support is detected per provider/model with a probe that is rejected before any token is generated, a floating per-model toggle turns it on, and attachments, auxiliary calls or unsupported routes fall back to the original adapter. Replaces the assistant-message renderer for all models while enabled.

Fold completed turns into their final conclusions, split thinking from prose with a divider, and jump between turns from a rail dockable left or right with line or dot marks, an optional ring and hover summaries; it can hide the host right-edge TurnNavigator, load earlier history, and open a pre-filled diagnostics issue.

Two-way WeChat (iLink) bridge: turn-end and approval-request push, in-chat approve/reject and message injection, persistent dedup and convergent long-reply chunking; channel layer extensible to other IMs.

Open source links from DeepSeek Harness conversations at the referenced line, review Git diffs beside the chat, inspect paused variables, and turn explanations into searchable Markdown memory cards. The workbench also shows Python class relationships (beta). Go and Java navigation require language servers, and breakpoint debugging requires a debug adapter.

Per-session execution gears for the DSH composer: an OVERDRIVE toggle turns Turbo on for the current session only, the Fast and Max gears swap the agent's planning and verification policy and may lower the request's reasoning effort, and the Host keeps the authoritative mode plus per-turn latency, step count, tool calls and whether a verification command ran.

Per-message delete for DSH: remove a user message, one reply step, or a whole assistant reply from the derived model context through the official surface-replace contract, and hide it from the visible transcript without rewriting the append-only log.

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.

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. Within a turn it supports skeletonized information compression with a low original-text retrieval rate; at the tail of a turn it supports a turn-tail intent summary implemented purely by content substitution; context content stays safe, and the whole process runs inside the official JSONL compression.

Exports the agent loop to GreptimeDB as OpenTelemetry traces, metrics and logs, with seven Grafana dashboards covering token usage, cost, session outcomes and per-turn latency.

Embeds a real VSCode editor (code-server, or your local desktop VS Code via the bundled bridge extension) as a GUI tab: agent file edits open turn-scoped red/green diffs that survive tab closes and page reloads, files lock while the agent writes, and a watchdog supervisor keeps zero orphan processes.

Playwright/CDP web-fetch provider for the built-in web_fetch tool: renders pages in a real browser, denoises them (Readability + DOMPurify) and returns markdown.

Compact right-edge turn rail with hover summaries and click-to-jump navigation. Ships on npm as `dsh-turn-scrubber`.

Cost tracking priced from an online catalogue (models.dev + OpenRouter, 8000+ models, each call priced at its own instant): a per-turn cost line, a Cost tab attributing spend to tool output / model output / system prompt / commands, budget, monthly forecast, ~166 live-rate currencies, and a bill_stats agent tool.

Replacement web interface for the unmodified official DSH release: collapsible reasoning groups, a turn timeline with adjustable reading width, cards for tool calls, a model and reasoning-effort picker, an Agents view that shows subagents as a tree, optional chart and diagram widgets in answers, light and dark themes, and a phone layout. Installs from npm; uninstalling restores the official interface. An optional jev_presentation_advice tool, enabled only once you save a TypeSafe key, sends up to four recent user messages and a short outline of the planned answer to api.typesafe.ai.

Wenyuan Pavilion: an AI Team workbench for DSH with owned tasks, independent acceptance, an append-only ledger and concurrent projects. Install it into its own profile (dsh plugin --profile ai-team add @suifei/dsh-ai-team), not web or desktop: its patch sets that profile's sandbox to danger-full-access and approvals to never, and the member agents run with full access. The project's PMO acts automatically by default (every 60 s it resolves pending decisions and wakes owners to dispatch the next tasks); turn off PMO auto to send decisions to your inbox.

Six-layer local memory for dsh: captures turns, stages them through an L2 buffer into DL3 summaries via a staging model, injects a stable memory block each turn and replaces chat history with placeholders. Ships five memory tools and a config page.

Reports which skills each turn invoked (loaded by the model, typed as /name, or none) and manages the local skill directory: install from a pasted repository, folder, SKILL.md or zip address, set a Chinese display name in the skill meta.yaml, and remove a skill while keeping a backup.

Smart session titles for DSH: after each conversation turn an independent auxiliary LLM call summarizes the full user+assistant transcript into a title that follows the session's real topic instead of echoing the first message; the first message is titled instantly, failed built-in titles auto-retry on later turns, manual renames are never overwritten, and subagent/fork sessions are skipped.