user-client and admin-client are on different major versions of react-query, zustand, and mui-icons

[iamtanuj18]

Background

While setting up locally I noticed user-client and admin-client each carry their own node_modules folders (177 MB and 58 MB) while backend, common, and integrations do not.

The cause is that the two clients depend on different major versions of a few shared packages:

Package	user-client	admin-client
@tanstack/react-query	^5.56.2	^4.10.1
zustand	^4.5.5	^5.0.0
@mui/icons-material	^6.0.2	^5.8.3

Yarn (with the node-modules linker declared in .yarnrc.yml) can only hoist one version of a given package to root. When workspaces disagree on the major version, one wins the root spot and the other gets a local copy in its own node_modules.

Nothing is broken today. Each workspace's imports resolve to the correct version via node's module resolution, and the two clients never share objects at runtime.

Problem

The two clients already share a substantial base: React 18, @mui/material v6.1.5, @emotion/react / @emotion/styled, react-hook-form v7, react-router-dom v6, immer v10. The three drifted packages sit directly on top of this shared foundation. Aligning them would give three concrete benefits:

• Maintenance surface will keep growing. React Query v4 is essentially in maintenance mode. Security patches will taper off, adding to the same package-vulnerability churn flagged in Review Unused Dependencies #872.
• Blocks shared React code in common. Because the underlying stack is already aligned, any shared React component, hook, or state helper could be extracted into common for reuse across both frontends. The current drift on these three specific packages blocks that.
• API knowledge overhead. Anyone maintaining both frontends has to remember two API generations per library. React Query v4 → v5 in particular has meaningfully different syntax (object-syntax hooks, placeholderData replacing keepPreviousData, changed filter callbacks).

Proposed approach

Rather than a big-bang alignment project, do the migrations opportunistically as each area is touched. Ordered from lowest effort:

• Bump @mui/icons-material in admin-client from v5 to v6. Trivial (icons are SVG components, near-zero API surface). Bonus: admin-client is already on @mui/material v6.1.5, so the current v5 icons combo is arguably an inconsistency worth fixing anyway.
• Bump zustand in user-client from v4 to v5 next time someone touches user-client's state layer. Largely backward compatible.
• Bump @tanstack/react-query in admin-client from v4 to v5 next time someone touches admin-client's data-fetching layer. Non-trivial — every useQuery / useMutation call needs review.

Alternatively, if the team prefers to knock all three out as one small tech-debt task, that also works.