omm.run

The bilingual official website and command reference for OMM. It is a Next.js application deployed to Cloudflare Workers through OpenNext.

Local development

npm ci
npm run dev

Open http://localhost:3000. English is served without a prefix and Korean is served under /ko.

next dev does not initialize Cloudflare bindings, so local development does not require Cloudflare authentication or invoke remote Workers AI. The assistant uses its deterministic fallback without the live inference configuration.

OMM AI assistant

/assistant and /ko/assistant provide a constrained OMM command selector. Cloudflare Workers AI can choose only an allowlisted command ID; all commands, options, examples, risks, and links are rendered from the existing static OMM command docs.

The Workers AI binding is declared in wrangler.jsonc. Live inference also requires an approved D1 ASSISTANT_DB binding and a server-side ASSISTANT_HASH_SALT; without both, the route fails closed to deterministic search. See the backend contract before configuring or deploying it.

Verification

npm test
npm run lint
npx tsc --noEmit
npm run build

npm run build runs the Next.js build once through OpenNext and produces the Cloudflare Worker bundle in .open-next/.

Pull requests and pushes to main run tests, lint, type checking and the Worker build in GitHub Actions. The workflow uses no Cloudflare credentials and does not deploy or call Workers AI.

The navigation badge reads the published omm-model version from PyPI. It is hidden if that request fails; the development version on GitHub main is not presented as a release. The footer commit is embedded at build time from Cloudflare Workers/Pages, GitHub Actions, Vercel, or the local Git checkout.

To compare the website command manifest against a trusted current local OMM checkout without network access:

OMM_SOURCE_DIR=/path/to/omm npm run check:omm-sync

npm run deploy changes the Cloudflare account and production Worker. Do not run it without explicit deployment approval.