Describe flow design
Lead: The build is finished and in two open PRs. Nothing is live yet, because a merge is a production deploy and needs your go-ahead.
The real panel, desktop
This is the real component in headless Chrome with real 3D. It is on a test page, not in the full app. Its height is 336 px.
The real panel, phone
At 390 px wide, the height is 573 px and nothing overflows sideways. Family and state rows swipe.
What a model drew from a sentence
19 of 20 descriptions gave a valid character on the first try. The pencil failed the eye check, and the app retries that case once. These ran on Haiku 5.5, which may not be the app's default Cloud model. The robot vacuum is only a blob.
What shipped in the PRs
| Area | Result || --- | --- || Library | 8 families and search, one family at a time || Describe | Model call with one retry, or copy prompt and paste || Build | Now part of Library; shapes, finishes and accessories kept || Originals, Minimals | Redrawn as outlines that take any colour; old bots keep their art || Colours and finishes | 11 colours plus custom; Flat is a seventh finish || States | 9 previews; happy and sad eyes on drawn characters || Phone and PWA | Same panel; accent colour moves under More options || Docs | Dev guide, user guide, mobile page, dependency map |Open limits
Check these first after the merge
A real signed-in Describe call has not run. The full app shell and vite build have not run. Check Describe once on ai.allternit.com after the merge.
- Yours lives in one browser. It does not follow a person to another device.
- Photos do not land in Yours, so a likeness is not stored in the browser.
bot_createdoes not set a look. Bots made from chat still need one picked by hand.