Appendix B — TanStack DB Deep Dive

Verified against @tanstack/react-start v1.168.x — July 2026.

Note

This is deliberately a short chapter, not a stub. TanStack DB is still 🟡 beta and has no SSR support today — in a Start app, it’s a client-only tool (see TanStack DB for the full derivation of what that means for loaders and hydration). Once SSR support lands, this appendix is the place that gets expanded into a real deep dive rather than a reference table.

For the concepts — collections, live queries, optimistic mutations, and the query-db-collection bridge to TanStack Query — go to TanStack DB. This page exists as a quick lookup for which collection type maps to which package, so you’re not re-deriving that from the main chapter every time.

Collection types

Collection	Package	What it syncs against
Query	@tanstack/query-db-collection	Any REST/GraphQL endpoint, via TanStack Query underneath — poll-based, not push. The bridge collection type; usually the first one you reach for.
Electric	@tanstack/electric-db-collection	Postgres, via ElectricSQL’s sync engine — real push-based sync, sub-2-second consistency in practice.
TrailBase	@tanstack/trailbase-db-collection	A TrailBase backend (SQLite-based sync backend).
RxDB	@tanstack/rxdb-db-collection	RxDB’s own local-first database, for apps already built on it.
PowerSync	@tanstack/powersync-db-collection	PowerSync’s sync engine, SQLite-based local-first sync.
Local	@tanstack/db (core, no separate package)	In-memory or localStorage-backed collections with no external sync target — useful for pure client state that still wants DB’s query/mutation ergonomics.

None of these ship a different mental model — a collection is a collection regardless of what’s behind it, and the live-query and optimistic-mutation API surface is the same across all of them. The difference is entirely in setup (the options object each factory takes) and in how fast/how consistent the sync actually is, which matters a lot for deciding what kind of data belongs in DB at all — see the decision framework for that call, and the trading/real-time pattern for a concrete case where DB’s sync latency (fine for positions, wrong for tick-by-tick prices) is the deciding factor.

Why this stays thin for now

Every collection type above is client-side sync — none of them currently have a story for participating in Start’s SSR loader/hydration cycle the way TanStack Query does via the official integration. That’s not a gap in this book; it’s the actual current state of TanStack DB. Once that changes, this appendix grows into the place with SSR-specific patterns, hydration guidance, and per-collection setup depth. Until then, the main chapter plus this table is the honest scope.