Feat: Semantic Tokens

[Mitch-At-Work]

Previous Behavior

Components relied on fluentui/tokens package which used generic level tokens for all components
fluentui/tokens was exported as a singular object, causing larger bundle sizes in ESBuild based projects

New Behavior

Components can now opt in to semantic style hooks via fluentui/semantic-style-hooks and fluentui/semantic-tokens package which:

• Enables semantic token specificity (Elevates fluent to the global Microsoft horizontal token definition).
• Provides token fallbacks to generic layers for minimal token definitions and theme definitions
• Is backwards compatible with existing token definitions, there should be no visual regressions when using existing token sets (original legacy tokens are the final fallback if no semantic tokens present)
• Provide additional tokens for use outside of fluent components, in a tree shakable manner
• Internally using flat exported versions of fluentui/tokens to ensure no increased bundle size when using ESBuild
• Move updated semantic styles to useStyle custom hooks so that semantic tokens are opt-in
• Move semantic-tokens to pre-release versioning (0.x.y)

ToDo (Post merge/preview release)

• Visual regression tests (For now, compared against master/legacy styles to ensure backwards compatibility).
• Storybook-addon: Enable opt-in to semantic tokens, inject via withFluentProvider, and enable example kumo theme via it's own @fluentui/semantic-theme package
• Semantic-tokens grouping update: Provide more fidelity/documentation around semantic token grouping, will extend/be backwards compatible with current tokens
• Finalize and add blocked components for full component coverage (Currently awaiting designs and token updates).

[github-actions]

📊 Bundle size report

✅ No changes found

[github-actions]

Pull request demo site: URL

[Mitch-At-Work]

Moved core changes (nessecary exports) to #35007

[Tudor Popa (tudorpopams)]

Closing this out as part of a review-queue cleanup: there has been no activity here for roughly 350 days, and the branch no longer merges cleanly against master.

This is not a rejection on merit — it's just housekeeping so the review queue reflects work that's actually in flight. If you're still planning to pick this up, please reopen it or open a fresh PR against master and we'll get it reviewed.