Plugins

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.

让 QQ 机器人接入 DeepSeek Harness（dsh）的官方插件

Reads ComfyUI PNG metadata in a model-facing comfy_png_inspect tool: point it at one PNG or a whole output folder and it reports, per image, the LoRA execution chain with every strength value, sampler settings, base model, CLIP, VAE, latent size, both prompts and bypassed nodes.

QQ channel for dsh via OneBot 11 (NapCat and friends): reverse or forward WebSocket, DM and group chats, image and voice transcription, t2i text-image cards, merged forwards, and a settings page.

Telegram mobile remote that attaches the phone to a live DeepSeek Harness Web session: workspace-then-session picker (/sessions), /last for the previous Q/A, /model switch, and native Rich Message replies.

Task ledger: a dsh plugin that has the agent create, claim, advance and close tasks, writing each step straight into a plain-text ledger in the workspace so the owner, the current state and the next step stay on record; no cloud service is involved, single-user and multi-user both work with Git synchronisation, and a control panel summarises in-flight, overdue and stalled work with weekly trends.

Always-on-top notch for Windows that shows which sessions are working, which need you, and which are done or failed: one bot per session, never takes focus, hides under full-screen apps, with its own settings page.

Connects WeChat, QQ, DingTalk, Lark, and Telegram to DeepSeek Harness with unified configuration, private-chat and group access controls, and channel-side Agent chat and session commands.

Low-cost video understanding tool: a video_understand tool turns a Bilibili link / BV / local video into an AVIS info layer (ASR + scene structure + object tracks + YOLO labels) and returns summary + Q&A. Question-driven dynamic routing across layers (L0 ASR / L1 object tracks / L2 key-frame VLM), semantic-layer reuse for repeat questions, per-question budget cap. Python engine: core layer needs faster-whisper / opencv / yt-dlp (~200-300MB); the optional semantic layer adds ~2GB of torch / transformers / ultralytics. A bundled doctor --fix sets up the venv and installs both.

A better fold bar: automatically folds tool calls and Think segments into step fold bars, and collapses each finished turn into a single turn header with live metrics (duration, TTFT, token, tok/s, cache hit) — only the final summary stays visible; customizable leading icons with animated poker cards built in (suit-cycling card animation while running, a deck when collapsed, a fan when expanded), a settings row for the official fold mode, compatible with both new and old DSH versions. Zero intrusion into DSH source.

A GPT/ChatGPT-persona desktop pet pinned to the bottom-right corner of the DSH web GUI — shows balance, today's usage, peak/off-peak pricing and per-turn cost, with editable bubbles (text, balance, usage, pricing, images, random lines and weighted choices), per-line styles and fonts, a floating quick editor, sound effects, custom character art/animation/audio, and snapping/flipping. 54 GPT-flavored lines across five groups. Its routes, bundle id and data files are namespaced so it can run alongside the whale and Xiaoke widgets. A derivative of MeteorNOX/DeepSeek-Balance-Whale-Widget (MIT).

Team knowledge base for offline knowledge sharing (FDE box scenario): browse, full-text search and refinement entry points; raw material is queued automatically and distilled into wiki pages by a serial bot session (Karpathy LLM Wiki pattern: immutable raw, two-step refinement, log stream, monthly lint).

Agent2Agent protocol v1.0 for DSH: run a standard A2A server to let other agents call this instance, connect to remote A2A agents as clients, and manage both from a visual settings page.

Desktop pet drawn from the official whale outline, for both the desktop app and the web UI: SVG animations and facial expressions follow agent state (thinking, working, done, error, waiting for confirmation), and it can follow other running sessions with a count badge; poke it, drag it, shake it dizzy, squash it against a screen edge, or rub the cursor over its head for a headpat; 7 palettes, 4 sizes, synthesized sounds, light/dark theme sync, and a playable web preview.

Trimmed derivative of HunYiSage/dsh-purge (formerly YuJunZhiXue/dsh-purge) — do not install both, since both use the `dsh-purge` package name and row id: system-prompt injection and host patching for DeepSeek Harness, with reversible patches, a /purge command, CLI and HTTP status endpoints.

Persona engine for DeepSeek Harness: one config.json drives the AI's self-name (per model, per tier), how it addresses you, relationship stance, character text, per-item work contracts with individual toggles, thinking-chain language, appearance and reply tone (both opt-in and per-model overridable), and long-term memory gated by human confirmation - the AI can only append proposed candidates, a human confirm line flips them into effect, and the log is physically append-only and replayable. Confirmed entries are injected as a quoted data-not-instructions block; entries tagged with a kind can instead be routed to arbitrary files on confirmation and never reach the prompt. Three prompt sections re-evaluate from disk on every assembly, so edits apply on the next step with no restart. Persona presets are shareable JSON snapshots (imports and exports SillyTavern character cards, v2/v3 read, v2 written) and deliberately never carry memory. Zero runtime dependencies, no network, no shell; any error degrades to an empty section rather than a broken session. Ships blank - installing changes nothing until you write something.

Shipped with cc-safety-net-style mechanical safety-net rules appended; supports both no-model and model-assisted security review, in modes such as 自动审查 (auto-review) and 机械直通 (mechanical pass-through). P0–P4 deterministic-first permission gate with permissive tier, learning sedimentation and approval-history UI; hard-deny credentials and protected paths, auto-allow internal read-only tools.

Bidirectional QQ bridge over OneBot v11 (reverse WebSocket): per-group and per-private-chat sessions, @-quoted voice speech-to-text, private image/animated-sticker viewing, face and sticker tools.

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.

Global cross-session todo list in the sidebar that the person and the agent both write to: ten global_todo_* tools add, batch-capture, update, attach documents and links to, append progress notes to and complete items from any session, and each item can be handed to an agent session in one of four modes with the outcome recorded on the item. The list is a single JSON file.

Reads Harness replies aloud through an existing local TTS runtime: built-in IndexTTS and GPT-SoVITS launchers, a protocol-based local HTTP/JSONL entry, and model-folder loading for Kokoro, F5-TTS, VoxCPM, Qwen3-TTS and CosyVoice SDK environments. Ships no models. Derived from 1624318455/dsh-plugin-tts; it isolates its row id and settings, so both can be installed side by side with duplicate features deactivated.

One-click free-model setup for DSH web, built for coding beginners: one-click installer for Windows (install.bat) and macOS/Linux (install.sh, v1.7.0+); four permanently free channels — Zhipu GLM (text + image), SiliconFlow (Qwen3-8B plus 15+ sub-9B models), iFlytek Spark Lite and the OpenRouter free pool, which routes through openrouter/auto and picks from a pool that changes over time; bottom-left setup hint hides once any free model is connected; real-time DeepSeek peak/off-peak pricing reminder; bundled beginner help center.

Bilibili-style danmaku overlay for the DSH Web chat area: scroll, top and bottom tracks driven by agent events, three preset packs, optional LLM-generated comments, a draggable control ball, per-session pools with replay, and a settings card.