How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

desktop - Skill - OpenSmartRoute

Imported from LemmaHQ/lemma (desktop/AGENTS.md). Install upstream with npx skills add LemmaHQ/lemma --skill desktop. Copyright stays with the author.

AGENTS.md — Desktop shell

English | 中文

These rules supplement the repo-wide conventions; the TypeScript rules apply here.

Engine and renderer

The renderer is the web client. npm run build builds web/ into web-dist/, then type-checks the desktop sources with tsc; change the UI in web/. main.ts serves web-dist/ under app://lemma/ and answers any path without a file extension with index.html, so the web client routes by path as it does in a browser. The window opens while the engine starts; connectEngine() waits for it, and a failed start replaces the page with the error. Git ignores web-dist/, out/, and .vite/.
src/engine.ts runs the engine in an Electron utility process. src/engine-host.ts is that process's entry: it loads the lemma-desktop addon, opens the engine over Electron's userData directory, and reports ready, or failed with the reason. Unpackaged builds load the addon from target/<profile>/, where LEMMA_ENGINE_PROFILE selects the profile (default debug); packaged builds load it from the resources directory.
The page talks to the engine over the engine port, a MessagePort that bypasses the main process. On connectEngine(), main.ts opens a MessageChannelMain and hands one end to the engine process and the other to the page; src/serve-engine-port.ts serves requests on the engine end. When the engine exits, main.ts tells the page, whose pending and later requests fail.
src/preload.ts implements window.lemmaDesktop, whose types live in @lemma/desktop-bridge; web/src/platform/ is its only consumer.
src/gen/theme-colors.ts carries the window colors: the background, sidebar, and foreground of each theme, generated from design/tokens.json. Git ignores it; npm run lint, test, and build generate it first. src/main.ts reads the titlebar and window background colors from it.
LEMMA_DEBUG_PORT enables Chromium remote debugging.

Commands

just desktop-dev        # build the engine addon, then launch Electron
just desktop-lint-check # oxlint; desktop-lint applies fixes
just desktop-fmt-check  # prettier --check; desktop-fmt formats
just desktop-test       # vitest
just desktop-cov        # vitest coverage
just desktop-package    # release engine addon, renderer build, electron-forge package

Install desktop and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

desktop

About

AGENTS.md — Desktop shell

Engine and renderer

Commands

Install desktop and it becomes one more candidate the router can pick - when it fits.

What it costs to adopt - in engineering time.

One base URL

No vendor lock-in

Failover built in

Use it

Manifest

Reviews