A2uiHost

The A2UI surfaces of one transcript, kept in an A2uiRenderer and drawn where the transcript's parts are.

Holds the A2uiSurfaces bookkeeping and applies each step to renderer as it comes. The slot table slots is what the drawing side reads: withA2ui looks up the carrier a part maps to, and draws the surfaces, the lifecycle, or nothing, as the reconciler decided.

A batch the renderer throws on -- a data-model path that is not a JSON Pointer, say -- is reported through onWarning, recorded as refused so it is not retried until its payload changes, and shown in its slot as the message the renderer gave. The other carriers' surfaces stand.

One host per transcript, and one renderer per host: A2uiSurfaces assumes it is the only thing creating surfaces in the renderer it plans for, and a second host sharing the renderer would delete the first one's surfaces as strangers. Which is why the three settings are vars and not construction-time: a host rebuilt for a new translation would start with empty books against a renderer that still holds every surface, and its first step would try to create what exists. Change the setting instead and accept again; the reconciler sees the payloads that changed and redraws those.

Parameters