Moonkale

ADR-0001 — Dioxus instead of Lumino + Tauri

1 min read

Status: accepted

Context

The original plan (rough goal) was Deno + Vite + Lumino (TS) for the frontend and Tauri + Rust for the backend. The brief now asks to stay in the Rust ecosystem as far as possible, serve desktop, web and mobile, and be extension-driven with a Rust-facing API.

Decision

Use Dioxus 0.7 for the whole application: one Rust codebase, three targets (desktop webview, web wasm, mobile), server functions for the backend. Lumino’s docking is replaced by dioxus-workbench (ADR-0010 dioxus-workbench for layout). TypeScript survives only as isolated view packages (ADR-0002 Rust first, TypeScript behind traits).

Consequences

  • One language, one type system across UI, model, sources and extensions — the extension API can be a Rust trait.
  • Mobile comes almost for free; with Tauri it was a separate effort.
  • Lose the TS ecosystem’s breadth; mitigated by the interop boundary.
  • Dioxus 0.7 is younger than Lumino; expect to contribute upstream (the workbench crate already exists because of this).
  • The desktop app is still a webview, which creates the graph-surface problem (ADR-0003 wgpu for graph rendering). The escape hatch (dioxus-native) requires no JS dependencies — see JS Interop Boundary.

Graph View

Backlinks