Plugins

链本体（三级模型：源 → key → 额度，按 key 记账/闸门/自动停用）：注册 web_search 的 provider 并供应配置卡片。由 dsh-websearch-tavily-keys-select 依赖，通常无需单独安装。

Auto-continues turns stalled by DSML tool-call markup leaking into reasoning.

DeepSeek Harness 情绪模块插件：感知/状态机/LLM复核/安全守卫/持久化/expressionStyle三档/对话框情绪chip开关（TypeScript + vitest）

Local tool guard that scores every DeepSeek Harness tool call with an Ollama decision model before it runs; risky calls come back as an approval prompt, and an unreachable model stops all calls. Requires DSH 0.2.x and a local Ollama 0.35.1+ with the default ~6 GB clef-flash model pulled first; without them every tool call is denied.

Codebase knowledge graphs for DeepSeek Harness: seven graphify_* Agent tools, a /graphify command, and a Graphify page build and query a project graph locally with AST parsing (no API key), plus path/explain/god-node lookups and wiki/HTML/SVG/GraphML/Obsidian exports. Requires the graphify CLI (graphifyy); only full document/PDF/image extraction needs an LLM key.

Adds the models of your own GPU Hugging Face Space to the model selector, starts the Space when you send a message, shows its start progress, and pauses it when idle or at a time limit. Requires a Hugging Face account with a GPU Space serving an OpenAI-compatible API, a Hugging Face token stored in the DSH vault as HF_TOKEN, and a paid plan for GPU hardware; Hugging Face bills the GPU time.

内置 provider / 模型显示名的全量数据库：载入时按环境里实际启用的 provider 与模型覆盖显示名，只影响 UI，不改路由、请求与凭据。

One-click polish, expand and autopilot for the DeepSeek Harness composer: rewrite the draft with the model the session is switched to, expand a one-line wish into a task brief, or let an AI drive a goal round after round.

Agent tools that download a Bilibili video, transcribe it with a local whisper.cpp build, then write a structured Markdown note with DSH's own LLM (optionally printing it to PDF). Windows only: missing dependencies are auto-downloaded into the data directory — yt-dlp.exe and the ffmpeg.exe build from GitHub and gyan.dev, the whisper.cpp x64 binary from GitHub, and a ~547 MB model from the hf-mirror mirror; PDF export shells out to a locally installed Edge/Chrome. Two tools: video_to_notes and note_to_pdf. Pure JS, no extra API key.

Deepseek harness plugin for control rate limits

Syncs the groups, keys and models of one sub2api instance into dsh-llm-pi-ai provider routes, with panel-driven login (password / OAuth) and one provider route per relay key.

See your real ChatGPT / Codex subscription quota inside DSH. Reads /wham/usage: 5-hour and weekly windows, reset countdown, extra-usage balance, additional limit groups and the code-review limit; lists and redeems server-issued reset credits with a stable, retry-safe idempotency key; switches between multiple accounts; reads /codex/models to show the account's real model catalog, showing empty rather than a built-in list when unavailable. One-click sign-in from Settings (browser authorization or device code, with a paste-the-callback fallback); credentials stay in DSH's credential service. Model calls reuse DSH's built-in llm-pi-ai Codex route instead of claiming a provider route — it degrades to a pure quota console when another plugin owns it. A remaining-quota chip sits beside the composer.

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.

Codex-style file-based long-term memory for DSH: two-stage LLM extraction, consolidated MEMORY.md handbook, prompt-injected memory summary, and search/read/list/note tools.

Limits the images sent to chosen providers or models in DeepSeek Harness: cap the longest side and the number of images per request, with its own plugin page

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.

Per-Session OpenRouter routing: a loopback relay writes `service_tier`, `provider.sort` and the `:floor` variant into each request from one of nine presets picked on the composer chip, with a pinned base URL per preset for route-level defaults.

Agent preset with adaptive tool presentation and environment-aware shell routing. Ports modules from yjh051108/dsh-routing-suite and xiaobright/dsh-anchored-standard (the former is listed here), so do not enable it alongside those presets. Requires DSH 0.2.2 or later; high uses PTC with on-demand tool unlocks and max uses PTC with the full available tool catalog.

Live2D character stage for the DeepSeek Harness Web UI: chat-column background, foreground overlay and sidebar seats, same-origin model serving, LLM emotion control with a cross-session director, SillyTavern-style character cards and world books, and a se

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.

OpenFOAM case doctor for DeepSeek-Harness. Six agent tools: audit a case for 25 classes of silent failure (steadyState paired with timeStep writes, limiters set to 1 that let k/epsilon go negative, unbounded advection of transport scalars, nNonOrthogonalCorrectors 0 on a skewed mesh, loose solver tolerances, blockMesh vertex/block mismatch, stray time directories, purgeWrite 0); read any dictionary as JSON; survey a sweep of cases; run checkMesh/blockMesh/snappyHexMesh/foamRun inside WSL with log parsing; parse logs into the metrics that matter (aspect ratio, non-orthogonality, skewness, residual floor, fatal errors). Deterministic text analysis — no LLM — validated against a real 82k-cell street-canyon snappy case.

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.

Audit what context compaction actually lost. Reads the raw session event log and checks session-constraint survival (the standing instructions the user gave — do-not-X, must-Y, only-in-Z — and whether the summary still carries them; arXiv 2608.11242 measured ~17% survival across compactors), summary hallucination (paths, versions, ports and identifiers the summary invents that were never in the history it replaced), ghost references ("see above" pointing at removed text), token accounting (a summary nearly as large as what it claims to free), and unaccounted prunes. Four agent tools including a cross-session sweep that ranks conversations by how much compaction hurt them. Deterministic — no LLM, so a wrong answer is a rule bug with a name.

Long-term memory on top of mem0 and a local Chroma store: user profile and recalled facts injected through the system prompt, capture at session end, a nightly pass that dedupes, merges paraphrases and resolves contradictions, a persistent topic graph written as Markdown cards, periodic export to a git repository, and a full-page memory UI with search, bulk merge, snapshots and undo. Needs a manual Python venv setup plus LLM and embedding API keys before use.