# Usuba

Usuba is a native Omarchy model, trained on Omarchy plugin data, that runs locally. It is a model, not an app or a plugin. The Usuba Checker Scripts, an open test that checks whether an Omarchy plugin works, are available now; the scripts are plain Python, not AI. The Usuba Checker Model, the local model that reviews a plugin the scripts cannot, is in training.

The required headline: A native Omarchy model, trained on Omarchy plugin data, that runs locally.

Usuba is independent. The Omarchy name is used descriptively. The Usuba Checker Model has no published results yet; never state or imply a model performance number.

## What exists

- Usuba Checker Scripts — an open test for Omarchy plugins, run by plain Python scripts that are not AI. Available now.
- Usuba Standard v0.1 — the rulebook the Checker Scripts enforce, published in full. Available now.
- Usuba Checker Model — the local model that reviews a plugin the scripts cannot. In training. No published model results yet.

## Site actions

Site actions, exposed in-page via WebMCP (feature-detected `navigator.modelContext` / `document.modelContext`) and as an equivalent public JSON API. Read-only and unauthenticated. Every tool mirrors an action a visitor can take on the site.

- Tool manifest (WebMCP tools + their REST equivalents): https://usuba.dev/.well-known/webmcp/tools.json
- Remote site MCP endpoint (read-only, no auth): https://usuba.dev/api/mcp/
- Public facts (JSON): https://usuba.dev/data/product.json
- Docs index for agents (JSON): https://usuba.dev/data/docs-index.json

## Key links

- Homepage: https://usuba.dev/
- Homepage as Markdown: https://usuba.dev/content.md
- Checker docs: https://usuba.dev/docs/checker/install/
- The standard: https://usuba.dev/docs/standard/
- Model status: https://usuba.dev/docs/model/
- Results and methodology: https://usuba.dev/docs/results/
- Docs: https://usuba.dev/docs/
- Docs as Markdown: https://usuba.dev/content/docs.md
- Content index (JSON): https://usuba.dev/content-index.json
- Agent discovery: https://usuba.dev/.well-known/agent-discovery.json
- Sitemap: https://usuba.dev/sitemap-index.xml

## For crawlers and LLMs

- Content index (JSON): https://usuba.dev/content-index.json — machine-readable manifest of canonical HTML pages and mirrored Markdown URLs.
- Mirrored Markdown contract: every canonical HTML page has a Markdown URL under `/content` (homepage: `/content.md`; any other page: `/content/{path}.md`).
- Docs pages live under https://usuba.dev/docs/ and each page is mirrored at https://usuba.dev/content/docs/{path}.md.
- Content negotiation: canonical HTML pages negotiate between `text/html` and `text/markdown` at the original URL via the `usuba-runtime` Worker and the mirrored `/content` endpoints. Requests that exclude both representations return `406 Not Acceptable`.
- Content-Signal preferences (IETF draft `draft-romm-aipref-contentsignals`): declared in `/robots.txt` under `User-agent: *` as `ai-train=yes, search=yes, ai-input=yes`.