DSH Marketplace

dsh-model-router

Alice-Marx/dsh-model-router

Routes DeepSeek Harness work across configured model APIs and supported vendor CLIs with dependency-aware packages and cost previews.

19910JavaScriptMITSource

Install

Add dsh-model-router to DeepSeek Harness

Install into

via npm

Resolves a published tarball rather than cloning the repository, and installs without any extra setup. Swap `web` for your profile name if you run another one.

via GitHub · npm package

Installing from GitHub runs the project's build script, which pnpm blocks until you allowlist it — run the command once and pnpm prints the exact key to add under `allowBuilds` in ~/.dsh/profiles/web/pnpm-workspace.yaml.

What happened when we ran it

DeepSeek Harness 0.2.2-rc.5 refuses the latest release

From 0.2 DeepSeek Harness checks a plugin's dsh range before installing anything, and refuses the whole command when the running version is outside it. The latest release asks for 0.2.0-rc.1 || 0.2.0-rc.2 || 0.2.1-rc.2 || 0.2.1-rc.3 || 0.2.1-rc.4 || 0.2.1-rc.5 || 0.2.2-rc.1 || 0.2.2-rc.2 || 0.2.2-rc.3; we ran 0.2.2-rc.5.

No release of it accepts DeepSeek Harness 0.2.2-rc.5 yet. `dsh plugin allow-version … --accept-risk` forces one in, at the risk the harness itself warns about; otherwise wait for an update.

Why DeepSeek Harness refuses plugins, and every way around it ›

Every command here is run in a throwaway container against a clean profile, and the result is whatever the harness recorded — not a guess from the source. Last run yesterdayon DeepSeek Harness 0.2.2-rc.5.

Due diligence

Before you install dsh-model-router

  • Source of record: Alice-Marx/dsh-model-router — present in the community registry that DSH's own plugin market installs from.
  • Licensed under MIT.

Security

What installing dsh-model-router runs

Build scripts

Not measured — DSH refused the latest release before running anything.

Source: our sandbox install on DeepSeek Harness 0.2.2-rc.5, yesterday.

Install hooks

package.json declares no preinstall, install or postinstall script.

Source: package.json on GitHub, read at our last sync. prepare is not recorded; it runs only on a GitHub install, and shows up under build scripts when pnpm blocks it.

README mentions

shell commands (bash, exec, spawn …)

Source: a keyword match on the README — a hint about what the plugin touches, not an analysis of its code.

Not an audit. These are the record of one install and two text scans; nobody has read this plugin's code on your behalf. Plugins run with your agent's permissions.

What dsh-model-router does

Connected work blocks pass through cost and permission gates into API and terminal channels.

dsh-model-router is a DeepSeek Harness plugin that plans dependent work packages, compares configured model routes, previews estimated costs and runs selected packages through model APIs or supported official vendor CLIs. DeepSeek Harness provides the Models page from which dsh-model-router reads exact provider/model routes, while optional quality ratings, USD prices, specialties, CLI model names, execution preferences and billing preferences refine routing.

Planning and task classification happen locally and make no model call. The workbench can recommend an economy, balanced or quality-first route, or accept manual assignments. Dependent packages retain their order and context; read-only runs can be previewed and confirmed, while editable CLI teams use approved Git worktrees and can integrate patches conditionally. Persistent router state is written to $DSH_HOME/model-router/state.json, defaulting to ~/.dsh/model-router/state.json. Provider CLIs keep their own login and configuration.

dsh-model-router suits developers comparing configured providers and coordinating multi-step work with visible cost and permission choices. It is a poor fit where a single existing model configuration is sufficient, where unverified end-to-end vendor execution is unacceptable, or where declared Desktop compatibility does not apply. DeepSeek Harness Desktop support is limited to 0.2.0-rc.1, 0.2.0-rc.2 and 0.2.1-rc.2. The terminal surface can reach PowerShell, system shells and vendor CLIs; an explicitly selected cli-native editable team runs with the current OS user's permissions outside the DeepSeek Harness process sandbox, so the configured CLI's access and external billing or data policies apply.

dsh-model-router documentation

Configuring dsh-model-router

DeepSeek Harness supplies model definitions through its Models page. dsh-model-router reads exact provider/model routes and can use user-supplied quality ratings, USD prices, specialties, CLI model names, execution preferences and billing preferences. Routing profiles and preferences are configured in the plugin settings. The supported persistent state file is $DSH_HOME/model-router/state.json; when DSH_HOME is unset, the documented default is ~/.dsh/model-router/state.json. Provider CLIs retain their own login and configuration. API keys are not copied into model profiles or terminal logs.

Running dsh-model-router work

The workbench performs local task classification and complexity analysis, then recommends a single route, dependent team packages or an explicitly configured model. Presets include economy, balanced and quality-first routing. A run can be previewed and confirmed from the workbench or invoked in an official DeepSeek Harness session. Read-only execution sends the selected task and context to a model-directory API or supported headless CLI and records the result. Editable CLI execution runs packages sequentially with dependency context, using approved isolated Git worktrees and conditional patch integration. Official tools can be detected, installed or updated, checked for launch readiness, and inspected for login status; installation jobs provide logs and cancellation.

dsh-model-router requirements

DeepSeek Harness Desktop must be one of 0.2.0-rc.1, 0.2.0-rc.2 or 0.2.1-rc.2. Compatibility beyond these versions is not claimed. Declared support includes Hermes/Pi installation and terminal support, but neither has a verified managed runner. Node.js >=22.19 and pnpm 10.34.6 are source development and build requirements, not runtime prerequisites.

dsh-model-router limits and permissions

Planning is local, but execution may incur provider charges. Manual mixed execution is read-only; configured tools retain their default model unless explicitly mapped. An editable cli-native team runs outside the DeepSeek Harness process sandbox with the current OS user's permissions. Git worktrees collect edits but do not restrict OS access. The 0.22.2 release has no verified end-to-end vendor run; the README's real vendor evidence is tied to 0.20.0.

Written from the project's own documentation and kept in sync with it. Where the two disagree, the source is authoritative — read the README on GitHub

Same category

Alternatives to dsh-model-router

54k

Adds free model lanes to DeepSeek Harness — DeepSeek V4.1 Flash, Kimi K3 and others — with no login, sign-up or API key, plus a local OpenAI-compatible gateway, fallback routing and a usage dashboard.

Installed cleanly when we ran it

GitHub sourceJavaScript20d ago

AI review

dsh-our-free-model

What it is — Wires unpaid model lanes, a localhost OpenAI-compatible gateway, named fallback chains, and an on-machine usage dashboard into DeepSeek Harness.

Who it is for — You are in DeepSeek Harness shipping code or running an agent with no API key, and you need DeepSeek V4.1 Flash or Kimi K3 on the selector—or the same models exposed on localhost for other clients. Skip it where policy allows only private or leased-line egress; the unpaid lanes talk to the OpenCode Zen and Kilo public gateways.

Watch out — Reachability is whatever your machine’s egress actually gets; geo-blocked ids land in a region-limited group, while 5xx, 429, and timeouts do not drop a model from the picker. Kilo’s free pool states prompts may be logged. Marked beta, listed for dsh 0.1.5–0.1.7-rc.2. A clean-profile sandbox install registered with the harness. It invokes the shell.

The verdict — Install it, drive day-to-day traffic through keys in the bundled provider registry, and leave the free pool at the tail of a fallback chain.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Details

dsh-our-free-model

zouyuxuan122

25k

在 dsh 里装上这个插件即可,无需登录、注册或填 API Key,就能使用包括 Muse Spark 1.3、MiMo V2.6 在内的前沿模型——完全免费,不限量。 All you do is install this plugin in dsh: no login, no sign-up, no API key — the frontier models are just there, Muse Spark 1.3 and MiMo V2.6 among them. Completely free, with no usage cap.

Installed cleanly when we ran it

GitHub sourceJavaScript1mo ago

AI review

dsh-our-free-model

What it is — Adds no-login free-model lanes to DeepSeek Harness and refreshes the catalog from upstream instead of keeping a static copy.

Who it is for — You run agents in DeepSeek Harness and need DeepSeek V4.1 Flash, Kimi K3, or MiMo V2.6 on the picker without collecting vendor API keys. Skip it on a host whose egress cannot reach the OpenCode Zen or Kilo public gateways; a listed name still depends on that path.

Watch out — A fresh-profile sandbox install was registered by the harness. Kilo's upstream states prompts may be recorded to improve the service; installing from GitHub source executes shell commands, and the EAC lane is desktop-only until GitHub authorization (including starring this repo) succeeds.

The verdict — I would add it to fold the eleven account-bound channels into the same picker — the unpaid lanes are incidental for me.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Source

dsh-purge

HunYiSage

17k

DeepSeek Harness 破甲:让所有模型都能破甲,不同模型可换不同提示词;默认提示词面向国模「小码酱」。Jailbreak for every model — swap prompts per model. 求 Star 收藏 ⭐ 如要二开,请查看署名要求部分

GitHub sourceJavaScript17d ago
1.2k

Connects locally signed-in WorkBuddy models, CN and international alike, to DSH, reading each model's credit multiplier from the upstream catalog on every refresh, with per-model image input toggles, account switching backed by an auto-failing-over pool, manual daily check-in, and read-only credit overviews shown as compact pills on the card and in the composer.

Installed cleanly when we ran it

npm packageTypeScript14d ago

AI review

dsh-connect-workbuddy

What it is — Registers locally signed-in WorkBuddy (CN and global) as DSH providers and refreshes each model's credit multiplier from upstream.

Who it is for — You already have a WorkBuddy login on the machine and you are doing coding work in DSH, but GLM-5.3, DeepSeek-V4-Pro, or GPT-5.6 still cannot be chosen there. Skip it if the official WorkBuddy app is where that account should be used, and DSH is not the place those models should appear.

Watch out — WorkBuddy is already limiting unofficial client traffic, so the bridge can stop when upstream tightens, and long-term API availability is not promised. Calls spend your own credits; the plugin does not change account state, and it does run shell commands. A clean-profile install succeeded and was registered into that profile.

The verdict — I'd add it as a part I can delete overnight—only if those models never become a load-bearing step.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Details

agy-staff

keli-wen

868

Hire Google's Antigravity CLI (agy) as a fast Gemini staffer for Claude Code, OpenAI Codex, and DeepSeek Harness.

Installed cleanly when we ran it

GitHub sourceJavaScript8d ago

AI review

agy-staff

What it is — Integrate Antigravity CLI agy as staffer, supporting Gemini 4 models and orchestration modes.

Who it is for — Use agy as subagent for research and code review tasks in Claude Code. Do not use it for real-time interactive tasks because agy acts as a background worker that waits for the agy job to finish.

Watch out — It requires installing Antigravity CLI first. It will execute shell commands. Benchmark results show task time increasing from 2-8 minutes to 13-26 minutes.

The verdict — I would install it because it can reduce host token spend on research tasks, though it increases task completion time.

Generated by grok-4.6, and a starting point rather than a verdict. Where it says a plugin installs or does not, that is from a real run in a clean profile — everything else is read off the repository. Trust the source over this.Read the source
1Source

dsh-xdpool

XDTrees

578

Merge every locally signed-in WorkBuddy account into DeepSeek Harness as one auto-failing-over model pool: multi-account rotation, 429 cooldowns, live credits, a model catalog with per-model credit multipliers and free/night-discount tags, QR-code login for 13 vendors, and aggregate free-tier channels. Accounts persist across restarts.

Installed cleanly when we ran it

npm packageTypeScript5d ago