Plugins

Adds a shared Sup composer switch on Windows to start or gracefully stop a separate Super-agent process and release its memory. Requires a working DSH NEXT install with Node 24+, the standalone DSH CLI and an already-configured independent Super-agent profile (verified on DSH 0.2.1-rc.4/rc.5); the bundled bridge.js must also be registered in the Super-agent profile — installing the desktop package alone does not complete setup.

Adds an 'Upgrades' tab to the DSH Plugins page: npm update checks for all installed plugins, engines.dsh compatibility preflight before upgrading, and one-click in-place upgrades that bypass pnpm's 24-hour release-cooling gate. Also refreshes the npm dependency families pulled in by GitHub-sourced bundles, and checks the Harness itself against the latest npm release.

Notify-only update checker for the plugins in a DSH profile: classifies each dependency by its specifier and offers only releases the running DSH core can load, in a plugin_updates tool and an Updates sidebar tab (the tab needs dsh-better-sidebar; the tool and settings work without it).

Region screenshot for DSH without a hotkey (Windows only; capture runs through PowerShell + GDI). Hold the left mouse button for 550 ms, the page dims into a crosshair, drag a box, release: the crop is saved to ~/.dsh/longpress-shots and copied to the clipboard, and a preview card can send the same bytes into the composer (button, Ctrl+V, or drag) instead of hunting for a file dialog.

Obsidian-compatible Markdown notes in the DSH sidebar with search, tags, backlinks, graph, audio notes with transcription, a paint canvas, OCR (needs tesseract), client-side encryption and notes_* agent tools; the release tarball and the npm package carry a prebuilt Linux x64 daemon, other platforms build it from source with Go.

Named model routes for subagent delegations, configured from a settings page: each alias binds a subagent provider, model and optional reasoning effort, token cap, agent preset, tags, plus a "when to use" line the model reads to pick it, so a delegation names only the alias. Adds five tools (delegate, list routes, edit config, wait, release), shows the route a child actually ran on in its session header, lets each UI element be switched off, and keeps all configuration in one local JSON file instead of YAML.

Fully local push-to-talk voice input for macOS: hold a global hotkey, speak, release — transcription is pasted at your cursor. Local-first STT (Qwen3-ASR 0.6B 4bit on MLX/Metal by default, SenseVoice and mlx-whisper fallbacks), zero external APIs, launchd self-healing service and an 8-point diagnose script. The bundled skill lets the agent install, configure (Karabiner hotkey wiring, launchd, engine switching) and troubleshoot the whole toolchain for you.

Git workflow studio for DeepSeek Harness: pick a branch model, enforce Conventional Commits, review PRs by defect density, rewrite history safely, rescue lost commits with reflog, and version releases by actual behavior change.

The menu bar app itself is a separate AppKit build from the repo's app/ directory (swift build -c release on macOS); the plugin only provides the host bridge.

Plays a character entrance card beside the composer when you switch model: identity name, six scored axes (intelligence, coding, speed, stability, cost, context), quota hint and a dialogue line. 23 curated cards across 15 providers; models without a curated card fall back to a data card derived from public benchmarks, or a generic card when no benchmark data exists. A web client plugin with no dependencies and no hooks.

Local video color editing with camera-matched DJI Log LUTs, independent highlights and shadows, optional skin and selective color controls, frame previews, and 10-bit HEVC export. The universal tarball listed here ships no runtime: its first-launch "configure environment" button downloads a pinned platform runtime (Python 3.12 + FFmpeg 9, ~100 MB) over HTTPS, or point it at your own Python 3.10+ and FFmpeg/ffprobe 8+; the platform-suffixed release packages bundle that runtime.

Dormant-by-default MCP for dsh: one mcp_summon tool; summoning a server registers its tools as real native mcp__<server>__<tool> tools for the rest of the session, and release unregisters them and stops the process.

When a live session's context passes a threshold, append older messages to a per-conversation folder on disk, then release the oldest finished turns from the model-visible context; the chat transcript and the session log stay untouched.

Host plugin for the macOS menu bar companion: live DeepSeek Harness session state, Approve/Deny for tool approvals, stop the current turn, open the exact session. The native app is a free universal build in GitHub Releases.

Per-package automatic updates for the DeepSeek Harness plugins you choose. On a schedule it asks npm for the latest version of installed plugins, updates only the packages you check, and skips local link or file installs, pinned versions and packages with errors; upgrades stay inside the major version and skip prereleases by default. Updates go through the profile's own plugin manager, so the profile write lock, pnpm step, bundle-patch validation and hot reload are the host's, and a failed install puts the previous version back. Each cycle is appended to a JSON log.

Holds every final answer before delivery, has one independent minimal-context model call re-derive and check it, and releases only the verified text; the draft is never shown. Reviewed answers cost one extra model request and appear a few seconds later.

Two switches at the right of the Session header. The left one stops writes to your local git repository (git add, git commit); the right one stops uploads to a remote (git push, git remote, gh). Both start on, and switching a guard on never releases the o

Version-adaptive tooling for DeepSeek Harness: it reads the running DSH version from the plugin manager, mirrors app-boot plugin-compatibility rules (engines.dsh first, then @deepseek-ai/dsh* peers, prerelease-inclusive comparison, exact-version exemptions) to score every installed plugin, and statically flags source still using APIs renamed or deprecated between v0.1.6 and v0.2.1. Ships a conversation-view panel, an agent tool and HTTP state/scan/report routes. Read-only, offline.

Read-only DSH profile doctor: the seven config-tree checks (config-clobber, dead-patch, tool-collision, plugin-not-mounted, plugin-stale, entry diff, explain) plus eight ecosystem checks learned from real failures — pnpm store version conflicts that break `dsh plugin add`, patch insert names node_modules cannot resolve, broken junctions, duplicate-capability plugins, orphan backup dirs, GitHub-source deps, npm publish-scope hints, and the minimumReleaseAgeExclude gap. Four agent tools; every finding carries its fix.

Reasoning-effort slider for the DeepSeek Harness composer: drag through the thinking levels the current model declares and release to write the chosen level into the session. An auto band escalates one level after a tool error and another after two or more, and optional model routing picks a provider/model per task type.

DSH plugin: one agent tool that audits a project for release-time infringement risk (licence contagion, copied third-party code, assets, trademarks, secrets) before publishing.

Plugin advisor for DeepSeek Harness: describe what you need and it ranks the npm packages carrying the dsh-plugin keyword by denoised metrics (daily baseline, week-over-week growth, release cadence, freshness) into a first pick, alternates and plugins to avoid; ships a sidebar panel, four AI tools and one-click install.

Restrict the file access of agents in any sandbox mode (read-only / workspace-write / danger-full-access / auto) to the session workspace, and ask the user to rule on cross-border access. Each session can also use the gear selection box next to the composer to relax the cross-border gate to 「 read-only outside workspace 」 or 「 fully release 」.

AI research toolkit for DSH: a model release radar, a per-model benchmark and community report from 13 structured leaderboards plus Chinese and English community discussion, and an AI news radar over 46 feeds and 22 official X accounts; registers four tools and three skills and renders every report to Markdown with a generated table of contents.