feat(registry): design-system registry infrastructure

[shokurov]

Summary

• Introduces a pluggable component + slot registry that lets consumers render the entire editor tree with any design system (MUI, Chakra, custom) without forking the library
• SchemaBuilderRegistryProvider nests — inner registries merge over outer, so you override only what you need
• Layout ownership enforced: DefaultFieldFrame and DefaultFieldHeader own their own visual classes; callers pass only semantic state (depth, expanded, hasErrors)
• Ships a full Material UI demo (demo/pages/MuiRegistry.tsx) that overrides every component adapter and slot adapter — demonstrates the API and justifies @mui/material as a dev dep

Public API additions

Export	Purpose
SchemaBuilderRegistryProvider	Context provider wrapping the editor tree
mergeRegistry(base, override)	Explicit registry merge helper
SchemaBuilderRegistry / SchemaBuilderComponents / SchemaBuilderSlots	Types
useComponent / useSlot / useSlotProps	Internal hooks (re-exported for advanced use)

Component adapters

Button, Input (with integrated label prop), Switch, Label, Badge, ButtonToggle (exposes aria-pressed for toggle state)

Slot adapters

FieldFrame (depth, expanded, hasErrors), FieldHeader, FieldMain, FieldActions, FieldBody, Root, MobileModeSwitch, FullscreenToggle

Test plan

• npm test — 4 new registry test files covering component overrides, slot overrides, mergeRegistry, and provider nesting
• npm run typecheck — no errors
• npm run check — no lint/format errors
• Open npm run demo → navigate to MUI Registry tab and verify all 6 schema fields render with MUI components

[lovasoa]

Hello and thank you! I find this a little hard to review. The pr is huge. Did you review it yourself?

[lovasoa]

The entire project has roughly 7k lines today, and this PR has almost 8k new lines and 2k deleted lines.

I think it's not fair to ask me to review all that, especially if you haven't reviewed it yourself.

jsonjoy-builder on  main [$] is 📦 v1.0.3 via  v26.3.0
❯ cloc src
      50 text files.
      50 unique files.
       0 files ignored.

github.com/AlDanial/cloc v 2.08  T=0.02 s (2034.1 files/s, 390180.1 lines/s)
-------------------------------------------------------------------------------
Language                     files          blank        comment           code
-------------------------------------------------------------------------------
TypeScript                      47            766           1023           7288
CSS                              1             62              7            269
JSON                             2              0              0            176
-------------------------------------------------------------------------------
SUM:                            50            828           1030           7733
-------------------------------------------------------------------------------

[lovasoa]

If you want to move forward, I think you should split this, and start with one small PR, that clearly brings value on its own, with a human written description of what you did to ensure the LLMs output is correct and why you think the changes are valuable, and don't introduce code duplication, unreachable hidden undocumented features, test cheats, illogical user experience, and other common issues introduced by unmonitored AI agents

[shokurov]

Hello, @lovasoa !

Of course I did review this PR. In fact this is 4th or 5th generation of internal reviews.
I've removed all the stuff that was non-essential, such as design docs, storybook and my internal CI/CD harness.

I think you are being unfair judging the size of the PR by sheer LOC.

In fact the only large pieces are:

1. the new demo page
2. test baseline snapshots (autogenerated)

Otherwise all the changes are atomic, clean extract refactorings of ui components. Even if I reduce the amount of these incremental changes (meaning less components migrate), that won't much affect total LOC, the demo will still be needed and the snapshots have to be updated.

I will be happy to assist with an overview doc, if you wish, to guide you through the changes, would that help?

Egor

[lovasoa]

Hi Egor!
I know you are motivated, and want this project to move forward. On the other side, I don't really have the resources right now to review large changes. What do you think about helping with maintenance yourself ? If you can promise you'll maintain this code in the long term, I can add you as an admin to this repo, and you'll be able to merge the PR yourself. You just have to commit to answer to potential future issues other users may have with it. Does that sound ok ?

[lovasoa]

And if you are in the mood of reviewing large changes, there is this other PR that sounds like a nice feature, but is also too large for me to review at the moment: #47 . Feel free to merge it if it's good.

[shokurov]

@lovasoa Yes, I can commit fair share of my time to project maintenance. Can't promise always day-to-day answers, though. What are your expectations? Are you up to continue doing maintenance?

[lovasoa]

Yes, I'm staying here as a maintainer. This is unpaid labor, so I won't require anything from you, but if you could at least review PRs and issues touching the code you added, it'd be nice. If you have more time and want to take more ownership, that's even better. I sent you an invite.

[lovasoa]

Publishing to npm is automated, so you should be able to publish new versions just by pushing a new tag: https://github.com/lovasoa/jsonjoy-builder/blob/main/.github/workflows/publish.yml

[shokurov]

Thank you, @lovasoa, I will have a look into #47 later this week