Tests and acceptance results

Run from the repository root with the locked dependencies restored. Read package.json for exact definitions; the table explains the scope rather than replacing those commands.

Command	Checks performed
npm run verify	Markdown, assets, script/UI logic, Svelte types, Rust fmt/clippy/tests and frontend production build.
npm run test:browser	Five-page browser interactions through simulated IPC; Windows uses installed Edge.
npm run build:desktop	Platform-specific Tauri release build and packages.
npm run test:headless	Real executable and SQLite with isolated synthetic sources; requires a build.
npm run test:desktop	Windows WebView2 and actual IPC; requires a build and an available CDP endpoint.
npm run test:import	Windows first GUI import at scale and full-process resource measurement.
npm run test:receiver	Windows actual IPC/HTTP/system credentials with owned-token cleanup.
npm run test:install:windows -- --help	Side-effect-free prerequisites for real NSIS lifecycle acceptance.
npm run test:install:linux -- --help	Side-effect-free prerequisites for isolated Podman package/GTK/WebKit/FUSE/Orca acceptance.
npm run test:credentials:linux	Native Linux Secret Service round trip on independent D-Bus/keyring storage.
npm run check:docs / build:docs / test:docs:browser	Documentation translations/types, production output/links and browser behavior.

Regression expectations

Ingestion changes should exercise meaningful input/output behavior, old summaries and consumed cursors, duplicate reads, conflicts, rollback, attribution and archive/retention boundaries. Unknown fields stay unknown. Default-zero test data without verified field validity must not become known zero. A source failure must not silently prevent other valid events.

Use a fresh test database per run and isolate source environment as well as app storage. Temporary scripts, logs, probes, extracted records and test databases belong under root build/<task>/. Do not reuse a PID-named directory that might reopen an old database.

Report accurately

Record cwd, command, environment/toolchain, exact code revision or working-tree status, exit code, time, first failure, retry result and scope. Keep static, synthetic, real-sample, mock-browser, native, installation, CI and production results distinct. A successful documentation build proves no application or provider behavior.

Use the validation template for new records. Plan.md owns active work; the current acceptance index links detailed records. Finish with git diff --check and inspect untracked files as well as the tracked diff.