API specifications are easy to write.
Governing them at scale is harder.

API specifications are useful only when people can find the right service and reach the right endpoint quickly. Raw files and scattered links make references difficult to search, compare, and keep close to the work.

YASP exists so a specification can enter a catalog, pass governance, and become an endpoint-centered working surface — without requiring a hosted documentation product.

Specifications can be large. Linting and analysis are expensive enough to freeze an editor if they share the UI thread.

Internal and pre-release APIs should not have to be uploaded to a third-party catalog. HTTP APIs and event-driven systems also use different description formats, so a single workspace has to hold OpenAPI and AsyncAPI without collapsing them into one model.

Live testing is part of documentation work, but the browser's same-origin rules and the risk of server-side request forgery both constrain how 'try it' can be implemented.

A document enters from a file or from an endpoint URL. YASP checks format and required metadata before the reference joins a collection. Collections stay searchable by name, version, description, and tag.

From there the reading surface is the endpoint or channel: method, parameters, and responses become the primary navigation. Optional AI-assisted generation and fixes sit at the edge of the system — they are not required for the core workspace.

The same core runs in the browser and in a Tauri desktop shell, so web and desktop are two surfaces on one catalog, not two products.

YASP is a Turborepo of three packages: a shared core for parsing, linting, storage, and AI integration; a React web app; and a Tauri desktop shell that wraps the same core.

The editor is CodeMirror. Spectral runs in a Web Worker. Collections persist in IndexedDB. A testing proxy stands in front of live requests so CORS does not define the product, and URL allowlisting plus private-network blocking keep that proxy from becoming an open relay.

AI-assisted generation and fixes go through OpenRouter when the operator opts in. Core documentation, search, and governance do not require a backend.

A specification becomes a searchable, endpoint-centered workspace instead of a file to open and scroll.

Governance, testing, and storage stay on the operator's machine unless an optional AI path is explicitly used.

Web and desktop share behavior because they share a core, not because two UIs were kept in sync by hand.

The interesting problem was not rendering OpenAPI. It was deciding which work belongs on the UI thread, which state must survive a refresh, and which network paths are allowed to exist at all.

YASP is a product-shaped system: catalog, governance, testing, and packaging. The stack is whatever those constraints required.