The desktop app now indexes through the shared provider adapters + persist
layer (scripts/providers/{claude,codex}, scripts/persist, scripts/parsing)
instead of maintaining its own parallel indexer. buildIndex shrinks from
~1173 to ~592 lines, eliminating the skill<->app parse duplication that
Phase 5 set out to remove. electron-vite bundles the .ts core from source
with better-sqlite3 injected; the provider->parsing graph stays
node:sqlite-free so nothing drags node:sqlite into the app.
Also fix a misleading log: when a manual rebuild tears down the worker
mid-build, the cancelled background build is a deliberate stop, not a
failure. Guard the service's failure log with the stopped flag so it no
longer prints "Obelisk index build failed: Indexer worker stopped" on
every rebuild.
- CONTEXT.md: provider-adapter + single-persist + node:sqlite-free parsing.
- docs/adr/0005: app builds with electron-vite (TS+ESM), packages with
electron-builder; preload CJS for sandbox; app consumes core from source.
Verified: full suite 121/121; a node:sqlite-adapter dogfood of the rebuild
path over real data (969 files, 285 sessions, FTS rebuilt) runs clean; app
Rebuild confirmed in real Electron/better-sqlite3.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2.5 KiB
The app builds with electron-vite (TS + ESM), packages with electron-builder
Context. The desktop app must consume the shared TypeScript/ESM Core
(providers/* + persist) instead of maintaining its own duplicate indexer, and
the app itself should be TypeScript + ESM long-term. The app previously ran raw
CommonJS on Electron's Node with only the Vue renderer built by Vite; the main
process had no build step, and Electron's bundled Node (20 on Electron 33) can
neither strip TypeScript nor use node:sqlite. Options for the main-process build
were a hand-rolled tsc/esbuild step, vite-plugin-electron, or electron-vite.
Decision. Adopt electron-vite to build all three processes (main, preload, renderer) as TypeScript + ESM, and keep electron-builder for packaging (dmg/nsis/AppImage). electron-vite is purpose-built for the Electron three-process model and handles the parts a DIY build would force us to hand-maintain forever (per-process module format, native-module externalization, dev reload). Specific decisions within this:
- Preload is emitted as CommonJS even though the app is ESM: the sandboxed
renderer (sandbox is on by default since Electron 20, and we keep it on for
security) does not support ESM preload. Source stays ESM; only the preload
output format is CJS.
mainloads../preload/index.js. - The app consumes the Core from source: electron-vite/rollup bundles
scripts/providers/*+scripts/persist(and theirscripts/parsing.mjsdependency) into the app's main/worker build, injectingbetter-sqlite3. This works because the provider→parsing import graph is node:sqlite-free (ADR-0001), so nothing dragsnode:sqliteinto the app. Thedist/frombuild:core(ADR-0003) remains for the skill artifact; the app does not need it. - better-sqlite3 stays the app's binding, externalized (not bundled) and unpacked from the asar.
Consequences. The app is restructured into src/{main,preload,renderer} with
electron.vite.config.ts; each main module is a build input so CommonJS-style
require resolution and the indexer worker ({ type: 'module' }) resolve at
runtime. npm run dev is electron-vite dev. Tests that loaded app modules moved
to ESM imports, and app-main-settings was rewritten from CJS Module._load
mocking to node:test mock.module (needs --experimental-test-module-mocks).
A future contributor may be tempted to make the preload ESM or disable the
sandbox — this ADR records that CJS preload under an on sandbox is the intended,
secure default.