Plugins

Auditable token-cost ledger for DeepSeek Harness: wraps the llm/stream waterfall and records every call with provider, model, input/output/cache-hit/cache-write tokens, usage source and integer nano-CNY cost from an explicit price table. Aggregates by hour, day and session; exports JSON or CSV with a no-write dry-run preview; keeps a manually entered official platform bill in a parallel column with actual-minus-estimate diff. Unknown provider/model routes stay explicitly unpriced. Zero runtime dependencies; no balance query, notifications or animations.

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.

English port of Windows Computer Use for DeepSeek Harness: 10 desktop tools (computer, computer_batch, open_application, windows, ui_elements, clipboard, switch_display, app_skill, request_access, list_granted_applications) with UI Automation, per-app access control, English settings UI and mini progress card overlay, Esc pause/resume, and DSH 0.2.1 unpacked koffi.node auto-discovery. Fork of copylee711/dsh-computer-use; it reuses the same plugin id and tool names, so the two cannot be installed together.

Standalone QQ-bot extension pack for DeepSeek Harness — fully independent, no other plugin required. Two-layer SQLite memory with hippocampal recall, a persona that keeps learning from daily chats, group tiers, local-model quick replies with a single-brain arbiter, TTS outbox, sticker library, an approval-gated video studio, and a four-form desktop pet (three custom characters plus the stock fish) playable over a public invite link with access-code login, single seat and brute-force protection. Chat bridging is deny-by-default: c2c/group access defaults to a care mode that only responds to the special-cared people and groups of the owner (shared with the persona system, zero extra config); allowlist, open and disabled modes remain available. Requires QQBOT_APPID and QQBOT_SECRET, and Node.js 22.5 or newer.

Shows today's, this month's and cumulative spend in the sidebar footer, under the message box and in the session header. Spend is metered by the plugin itself from session usage events using official DeepSeek unit prices (peak/off-peak, weekends and holidays treated as off-peak), with local-day aggregation and currency conversion. Also adds a cost_today tool and a /cost command.

Token usage workbench for token-plan users: folds assistant/message usage from the session log into per-day call-level detail (cross-session responseId dedupe keeps fork/resume copies counted once), converts local volume into plan credits against terms you enter (default 1M tokens = 0.2 credits, caps 600/2500/6000 for day/week/month, optionally scoped to selected channels), and shows a composer-dock readout plus a settings page with per-provider cost table, 7/30d bar and line trends and a rolling six-month heatmap. No credentials are read: the remaining percentage a provider keeps behind its control-plane API is replaced by a clearly labelled local conversion, and reference costs come from the public DeepSeek pricing page.

Adds a single “Restart DeepSeek Harness” item to the Windows system tray menu of the DSH desktop app by patching the installed app.asar in place and re-applying that patch after each desktop update; it needs a writable DSH installation directory and adds no other tray actions.

Pins an always-on-top progress bar while the agent runs a turn — the percentage is estimated from tool calls and filled to 100% when the turn ends — and badges the DeepSeek peak/off-peak pricing window so large token operations ask for confirmation only during peak hours. The bar runs as a separate Electron process and is skipped when no Electron binary is found.

Session status hub for DeepSeek Harness: a top-titlebar rail of running / waiting / unread / failed Sessions, Windows toast notifications, and a red-yellow-green taskbar status dot.

Power-Up — small, opt-in upgrades for DeepSeek Harness. Adds a right-click action menu to the Trajectory ledger, an optional stop reason in the session header's background jobs panel (appended to the injected completion notice), and an install control for

A Shenshen desktop companion for DeepSeek Harness: transparent task-aware animations (idle, working, thinking, celebrate, failed), click-to-wave and drag-to-run interactions, plus the upstream API balance, local usage accounting, roles, bubbles and sound settings. A character branch of MeteorNOX/DeepSeek-Balance-Whale-Widget (dsh-whale-widget 0.4.2); do not install it alongside the upstream, or two pets and duplicate requests appear.

Plugin Steward adds a plugin panel to the chat composer for managing the installed DeepSeek Harness plugins: list them, switch them on or off, edit a plugin's settings, hide one plugin's tools or MCP servers from a single conversation, and have the model write short Chinese labels for installed plugins by reading their own README and package files.

Forward WeChat chat records into DSH sessions: a Windows share target captures the merged-forward ZIP, an in-GUI overlay picks the destination workspace and session, and the ZIP arrives as a file-attachment user message the agent unpacks and reads. Media entry extensions are normalized by magic bytes (WeChat ships PNG/HEIC bytes named .jpg), and failed deliveries preserve the batch for retry. Requires Windows 10/11 with Developer Mode on and the repo installer to register the share target.

DSH Desktop workbench for official MATLAB MCP experiments, verified PNG figures, MATLAB-computed metrics, and PID parameter sweep comparisons. Requires Windows, a MATLAB license and the separately installed official MATLAB MCP Server.

Per-conversation behavioral rule fence for DeepSeek Harness: rules live in one JSON file, are edited from a Settings page or the conversation panel, and every rule carries its own on/off switch so a single rule can be muted without deleting it (muted rules are skipped at injection and the remaining numbering stays contiguous); any number of conversations can also be excluded as a whole so they receive no fence at all, and picking conversations never touches the rules themselves; the enabled rules are injected as one runtime-context section at a fixed order so the system prompt prefix stays stable.

Audit how each turn's context budget is spent. Breaks down what a turn pays for AGENTS.md, the skill catalog, tool schemas and runtime context; flags repeated injection of unchanged content as pure waste; ranks skill-catalog bloat (with 60+ skills installed, how many tokens of descriptions does the model read every turn) and tool-schema inflation per plugin. Deterministic — no LLM, so a wrong answer is a rule bug with a name. Four tools including a cross-session sweep.

Per-target scheduled wake-ups for a DSH workspace: interval, quiet hours, daily cap and a prompt per target; each wake-up waits for the turn it started and records the outcome, and a turn that never finished is not counted as done.

Thinking-sorting plugin. A switch beside the composer puts the session into a three-stage mode — sort, critique, research — and each stage has a checklist of what it should produce before you press the advance button; finished ideas are filed by genre into Settings → Thoughts Center. A floating notebook holds half-formed sparks and opens a new session with the chosen one already written into the input box.

Parent-agent dispatch with review: split a task into work orders, hand each to a contractor you picked (its own provider, model, and reasoning effort), then pass or fail every delivery — failed ones are re-dispatched with your feedback.

ZenTao work panel in the DSH left sidebar: lists tasks, bugs and stories assigned to you (web-login protocol, so instances where the RESTful v2 API is unavailable are supported — verified on ZenTao 18.13 open-source) with search, smart sort, overdue / due-today filters and a flow-board view; opens a new chat pre-filled with the card's description, recent comments and screenshots (or a blank one), or dispatches it to a new background session with workspace, model, reasoning-effort and optional-instruction selection (single or multi-select batch), turns screenshots inside ticket descriptions into real image attachments, reads back the linked session's latest reply once a card is done and lets you answer the agent from the panel while it waits on you, and tracks each card with live running / needs-decision / done status plus manual overrides and a manually-marked finished state, with a deep link to ZenTao's finish / resolve page. Multi-account; read-only, no writes back to ZenTao.

Lets the agent see the compiled page: renders any page to an image, resolves a PDF point or a text anchor to the exact source line through SyncTeX (and says when a rule is drawn by a package rather than by the body, which is where naive locating is wrong), diffs a page against the previous build so an edit is visible, audits every page for holes the edit introduced, and checks that numbered labels still read in order.

Image-generation tool with automatic prompt enhancement, switchable API profiles, and a Turn-tail image card that survives chat folding. Requires a user-supplied image API endpoint and key; none are bundled.

Reads a video or audio file into metadata, sampled stills, subtitles and a multilingual speech transcript with speaker labels, captures the live desktop as frames or a recording, and installs its ffmpeg/whisper.cpp/sherpa-onnx toolchain on first use with per-file sha256 checks; all processing stays local. Automatic toolchain install and desktop capture currently support Windows x64 only, so elsewhere ffmpeg and model paths must be set manually.

Keeps a local sticker dictionary keyed by content hash so an image already described is recalled instead of re-read, and splits a GIF into evenly sampled frames laid out as one contact sheet so the whole animation is visible.