Plugins

In-conversation source deck for DSH Desktop (supports DSH 0.2.2-rc.1): the current chat model runs LLM search rounds, candidates render as GitHub-aware cards, and keep/exclude feedback, favorites, and the selected sources stay in the same conversation.

A no-limit Texas holdem table where you are the only human. Eight seats: you plus seven LLM agents written on AI-founder personas (Altman, Dario Amodei, Musk, Liang Wenfeng, Jensen Huang, Lei Jun, Zhang Yiming). Each agent bets through the currently selected host model and falls back to its own heuristic style when no model is available; every agent only ever sees its own hole cards, because the prompt is built from that seat alone. A hand ends by revealing the winner hole cards, with a flip animation replayed on every community card, and a rail keeps a live action-and-table-talk timeline. LAN mode (off by default, single-machine play is unaffected) lets another install on the same network join this very table: the host shows a six-digit room code, UDP broadcast discovers nearby tables, and a joiner randomly takes over one bot seat, with the seat handed back to its character on leave or after twenty seconds of silence. The room code is never put on the wire and every snapshot is generated for one viewer seat, so nobody, not even the host, can read another player hole cards. Fork of BubblePtr/dsh-holdem (MIT), credited in the README and in the UI.

Long-term memory scoped to the DSH workspace. A session opens with a one-line-per-memory index - bodies are fetched on demand by the `memory` tool, so the same character budget holds about three times as many entries - and when a turn ends a purely local search over the last user message pulls the matching memories into the next briefing at no model cost. Optional auto-capture distils a stretch of conversation into structured memories; in draft mode they wait in an approval queue instead of landing directly. A portrait layer keeps a workspace and a global picture, rewritten from the journal and the records, with the same draft-or-direct choice. Settings → Memory holds the archive, list and timeline views, a health check (duplicate clusters, stale entries, broken links), the supersede version chain with rollback, and a read-only scan that imports existing WorkBuddy / Qoder memory files.

Runs the standalone opencode2api backend (jasonxu114514/opencode2api, Windows x64/ARM64, no Docker or OpenCode CLI) with DSH and on exit stops only the process it launched; its settings card installs, updates or rolls back the backend from checksum-verified GitHub releases, copies the initial admin password and syncs /v1/models into an opencode2api provider with context, vision and reasoning efforts, re-syncing every 10 minutes.

Google Antigravity and Cloud Code Assist model provider with Gemini 3.8 Flash, sandbox failover, and multi-account quota management.

Agnes AI provider for llm-pi-ai: 4 multimodal chat models declared in the bundle patch, plus image and video generation tools.

Reads each third-party endpoint's own model listing and keeps llm-pi-ai's per-model reasoningEfforts in sync, so the composer offers the thinking levels a model actually has.

Use Zed AI hosted models in DSH.

Model picker dialog with routing groups: a model that keeps failing hands the turn to the next usable model in its group, and the dialog also carries a metrics dashboard and a free-channel tab. That tab installs six channel plugins — Our Free Model (the only one that needs no login or API key), CodeArts, WorkBuddy, Trae, Qoder and Providers Hub — whose providers cover 23 platforms: Huawei Cloud CodeArts, Tencent CodeBuddy, WorkBuddy, Youdao LobsterAI, Qoder, TRAE, Cline, iFlytek Loomy, SenseTime Raccoon, MiniMax, Zhipu ZCode, Gemini Code Assist, Alibaba Accio, iFlytek AStudio, OpenCode, OfficeAce, AutoClaw, ModelScope, Qiniu, Codely, MonkeyCode, Baizhi Cloud and MagiCrew.

Registers external ACP (Agent Client Protocol) agents such as Codex, Claude Code and Gemini CLI as DSH model providers: maps each agent's thought_level options to the reasoning-effort selector, forwards images when the agent advertises promptCapabilities.image, and routes permission requests to DSH's approval dialog; ships an ACP Agents settings page that imports Zed agent_servers JSON and discovers models.

Runs a local OpenAI-compatible gateway so DSH can use the model quota of a logged-in Tencent CodeBuddy (WorkBuddy) account: one-shot installer with SHA256 verification, logon scheduled task with duplicate-launch guard, manage.ps1 for status/start/stop/restart/logs/update/uninstall, and an agent skill that drives setup and troubleshooting.

Registers an apiCost session projection that estimates DeepSeek spend from reported token usage at published peak / off-peak rates, in CNY and USD, over self, delegation-tree or verified-team scopes: a composer-dock pill and its detail panel, the /cost command, a session_cost tool, and GET routes for the view and the current rate card. A panel switch parks model requests at the llm/stream waterfall while peak rates apply and dispatches them unchanged once the off-peak window opens, so the call is billed at the off-peak rate with no failure reported to the session.

One-tap connect ModelScope language and image-generation models into DSH (off by default), with image recognition and context window filled in automatically. You can also see the free Magicube balance, jump to a model's info page, and record local call counts, trends and rate-limit events, with a usage pill beside the input box showing the balance and today's local count. All of it needs only an API key, with no account sign-in.

Composer prompt enhancement for DeepSeek Harness: a button beside the model selector rewrites the draft through one auxiliary LLM call and writes the result back in place, with composer undo left intact; a /prompt-enhance slash command returns the rewritten prompt as a command result line without touching the draft; the enhancement call can be pinned to its own fast provider and model in the profile patch; the rewrite policy is faithful - original intent, language and information are kept, code, commands and paths are preserved verbatim, and nothing is invented or padded with placeholder text.

ChatGPT-desktop-style model picker whose thinking-level slider carries a particle trail, plus a visual-only fast-mode toggle and a hold-to-lock for it.

Binds a Feishu/Lark bot to DeepSeek Harness over a WebSocket long connection: incoming messages run through an agent session and stream back to the chat, chat commands such as /attach manage the live session, and feishu_* tools let the model push messages.

Run one deterministic skill-routing pass on every user instruction, so the skill catalog is consulted before the model answers instead of after it remembers. On the host it scores the live catalog from the instruction's keywords and the catalog metadata, with no LLM or embedding call, and appends a single producer-tagged message to the step: a ranked shortlist by default, the winning SKILL.md body once the score clears a threshold, or an instruction to call a named router skill exactly once. Unrelated instructions stay silent.

Adds AMD Radeon Cloud as a model provider in DeepSeek Harness, with reasoning-effort levels declared per model. The provider definition ships in the plugin's own patch layer rather than in your profile.

Fills missing model parameters from models.dev, imports models from configured providers, and writes per-field overrides for built-in models, with prices in USD or native CNY set as manual rates or a reference price times a multiplier.

Rewrites drafts with one model call or predicts the next reply when empty; independent tasks, safe draft replacement, result previews, and original-text recovery.

Prompt optimizer for the composer: a button in the tool row rewrites your draft through a model channel you already configured, streaming into a panel that floats above the composer without taking up layout. Your draft changes only when you click Accept — and Accept is refused while the answer is still streaming, so a half-finished reply can never be written back. It asks first when the draft carries @ references or attachments, which a whole-draft replace would flatten into plain text. No global keyboard shortcuts, and no API key ever reaches the browser: the host half makes the call.

Browser provider for DeepSeek Harness: hand it a web task — open a page, click, type, fill a form, read what is on it — and it drives a real browser step by step, seeing an indexed table of what can be done on the current page instead of raw HTML or screenshots. Each step is one bundled-Jev decision rather than a full model turn (1.1 s median per step versus 2.8–3.6 s for chat-model turns in the repo benchmark), and the main conversation receives only the conclusion, so its context does not grow with the page. Targets come from the page it just read and a changed page is re-read rather than clicked blind, and it runs in the browser that Session owns, leaving your everyday Chrome untouched unless you switch to attach mode.

Cross-session message forwarding between DSH sessions: extract one model reply, a user message, or a multi-selected batch of messages and images from one session and deliver it into another, with silent send, draft fill, or send-and-jump delivery, plus prompt templates.

Shows what each completed turn cost in CNY on its own row under the turn, with a click-to-open breakdown naming the tokens per category, the unit rate, the matched price rule and the billing instant; per-model rates are edited on a settings page inside DSH — four token categories plus weekly recurring peak/valley rules with arbitrary multipliers and absolute time-window promo prices that override them. 0.2.3 adds account awareness: it reads the balance or remaining quota of 14 providers (DeepSeek, OpenRouter, Zhipu GLM, Moonshot, MiniMax, Kimi Coding, OpenAI-compatible relays and more) and shows today / this month / this session spend next to it; spend is accumulated in a local ledger (atomic writes, one book per API key, 1-24 month retention, JSON export), and a monthly calendar lets any day be corrected by hand - including days with no balance observation - with an optional note. Usage and rates are computed locally from what the provider reported; balance is read straight from the provider API with your own key, and nothing is sent to a third party.