Knovum: Local-First Desktop AI Workspace Powered by @tanstack/vue-query & Tauri 2

[madebyzan]

Hi TanStack community! 👋

We wanted to share an interesting non-browser use case for @tanstack/vue-query in Knovum — an open-source, local-first desktop AI knowledge workspace built with Vue 3, Rust, and Tauri 2.

Why TanStack Query in a Local Desktop App?

Most people associate TanStack Query with HTTP/REST fetching. In Knovum, we have zero cloud servers — everything runs locally on SQLite, embedded LanceDB, and on-device models.

Yet, @tanstack/vue-query is the backbone of our desktop data orchestration over Tauri 2 IPC:

1. Async IPC Bridge Caching: Communicating with our compiled Rust backend (e.g. querying vector embeddings, fetching citation metadata, or polling long-running Whisper transcription tasks) is inherently asynchronous. TanStack Query treats Rust IPC invocations just like server queries, providing deduplication, background caching, and automatic garbage collection.
2. Deterministic Cache Invalidation: When a user creates a new claim, links two notes on the 2D canvas, or updates an asset, calling queryClient.invalidateQueries() instantly updates all dependent canvas cards, relation inspectors, and outline panels without manual pub/sub spaghetti.
3. Optimistic Desktop Updates: Moving or grouping cards on the infinite canvas reflects immediately in the UI while the background Rust write commits asynchronously to SQLite WAL.

Combined with Tauri 2 native OS WebViews, our Vue 3 + TanStack Query frontend idles at only ~45MB RAM on macOS and Windows.

Huge appreciation to Tanner Linsley and the TanStack maintainers for building such a robust, runtime-agnostic state orchestration engine!

🔗 GitHub: https://github.com/madebyzan/knovum