Official codemod (run in a clean worktree; docker-owned data/ blocks it
in-tree): tailwind.config.js inlined into client/index.css as CSS-first
@theme/@utility (config file deleted — v4 auto-detects content, which
also retires the .tsx-content-glob failure class from the TS migration),
33 template files' class renames applied, tailwindcss-animate loads via
the v4 @plugin bridge. Build runs through @tailwindcss/vite; postcss
config is now plugin-free. tailwind-merge -> 3 (v4 class model).
Visual-regression gate: only diffs were a ~2px global preflight shift +
the version string (baseline predated 0.41.1); re-baselined both login
screenshots in this commit after review. webauthn.probe confined to the
probe project (it enables WebAuthn for the seeded user, racing the
parallel UI projects' password logins).
e2e 27 passed, probe 18/18, PROD SMOKE PASS, typecheck/lint/tests green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Build drops 6.5s -> 0.65s. Three breakages found by the prod-smoke gate,
all fixed:
- rolldown removed object-form manualChunks -> function form in
vite.config.mjs (same vendor split, all @radix/@tanstack grouped)
- the 4a 'send' bump rejects bare absolute paths: SPA fallback sendFile
and both download routes (admin backup, user-db export) now use the
root-option form
- rolldown minifies inline HTML scripts, breaking any CSP hash approach:
the pre-paint theme script moved to client/public/theme-init.js so
script-src 'self' holds with no inline hash to maintain
@testing-library/dom added (peer no longer auto-installed). Server 252,
client 52, probe 18/18, PROD SMOKE: PASS.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
eslint 10's recommended set adds no-useless-assignment; fixed the three
real dead initial assignments it caught (version.cts updatedAt,
simplefinService claimUrl/accessUrl). Lint 0 errors; ratchet rules from
the standards batch carry over unchanged. Server 252, client 52.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- nodemailer 8 -> 9 (GHSA-p6gq-j5cr-w38f, raw-option file-read/SSRF; we
never use the raw option — notificationDelivery tests 13/13)
- xlsx -> official SheetJS dist 0.20.3 (prototype pollution
GHSA-4r6h-8v6p-xvw6 + ReDoS GHSA-5pgg-2g8v-p4x9; npm registry line is
abandoned at 0.18.5 with no fix). API-identical for our 8 call sites;
import/export round-trip tests green. NOTE: the lockfile now pins
https://cdn.sheetjs.com/xlsx-0.20.3/xlsx-0.20.3.tgz — CI runners and
Docker builds need egress to cdn.sheetjs.com
- CI: npm audit --audit-level critical -> high (0 vulns as of this
commit; highs can no longer pass silently — QA-B0-02)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
De-risks the flagship. The app is synchronous (better-sqlite3) but Kysely
executes async, so db/kysely.cts uses Kysely as a TYPE-SAFE BUILDER only:
build the query (columns + user_id/deleted_at scoping + types checked at
compile time), `.compile()`, then run via sync helpers (all/get/run) on the
shared connection. No async ripple, no change to the sync model.
- Schema typed for `categories` (extend as adopted).
- tests/kyselyPoc.test.js proves byte-identical rows vs the raw SQL and
that compiled SQL is parameterized (injection-safe).
- Pattern documented in CODE_STANDARDS; adopt for new/refactored queries;
it type-enforces user-scoping (the biggest data-isolation win) and
removes Row=any at the call site.
typecheck:server + check:server clean; 248/248 server tests.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A financial app must prove its money math, not sample it. tests/
moneyProperties.test.js fuzzes utils/money.mts: toCents always integer;
dollars→cents→dollars == roundMoney (no drift); sumMoney is cent-exact;
integer cents round-trip; classic float-drift cases exact; invalid inputs
never yield a bogus number. 6 properties, all hold.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`npm audit fix`: react-router/react-router-dom 6.30.3 → 6.30.4 (moderate
open-redirect via protocol-relative URL), nodemailer 8.0.9 → 8.0.11.
Verified: server 232/232, client 50/50, build green.
Remaining audit findings are assessed non-actionable/mitigated (see
SECURITY.md): nodemailer `raw` send option is not used; xlsx is confined
to the import service with raw:false + size/content-type limits.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Pillar A enforcement so standards can't drift:
- .forgejo CI now runs format:check + lint (client+server) + typecheck
(client) + typecheck:server, in addition to the existing check:server /
tests / build. Previously lint and typecheck were NOT run in CI.
- lefthook pre-commit auto-formats staged files (prettier) + lints;
pre-push typechecks. Auto-installed via the `prepare` script.
- Knip config (advisory `npm run knip`) for dead-code/unused-export
review; the design-system ui/** is ignored to keep signal clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Standardization foundation (Pillar A):
- Prettier (.prettierrc.json, .prettierignore) + `format`/`format:check`
scripts; style tuned to the house style (single quotes, semicolons,
2-space, printWidth 100).
- .editorconfig for editor-level consistency.
- Extend the flat ESLint config to lint server `.cts` (Node globals, TS
parser, "focused signal" rule set mirroring the client) so `eslint .`
now covers both halves; `lint` → `eslint .`.
- Wire `format:check` into the `ci` script.
`npm run lint` is green (0 errors; pre-existing intentional patterns kept
as advisory warnings to ratchet later).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The lockfile had drifted (written by local npm 11 / Node 25) into a form that
node:22-alpine's npm 10.9.8 reads as incomplete — `npm ci` failed with
"Missing: esbuild@0.28.1 from lock file", breaking the Docker image build
(Dockerfile now uses `npm ci`, not `npm install`). Regenerated the lock inside
node:22-alpine with its own npm so `npm ci` resolves cleanly (887 packages,
lockfileVersion 3). No dependency version changes — lockfile metadata only.
Verified: clean `npm ci` in node:22-alpine, full image build, container boots
migrations + workers, full QA (auth/CSRF/admin/security) and a live SimpleFIN
sync all pass.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Delete unreachable in-file dialogs: MonthlyStateDialog block (setShowMbs
never called) and inline PaymentModal in TrackerRow + MobileTrackerRow
(setEditPayment only ever null; editing works via PaymentLedgerDialog).
- Delete fully-unreferenced files: tracker/EditableCell.tsx,
ui/confirm-dialog.tsx, ui/separator.tsx (+ orphaned @radix-ui/react-separator).
- Simplify OverdueCommandCenter dead branch (early-returns null when empty).
- Correct stale roadmap/FUTURE docs: tracker keyboard nav is implemented.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Establishes the server-TypeScript foundation and migrates utils/money →
utils/money.mts with branded Cents/Dollars (mirroring the client), so the
cents↔dollars boundary — the origin of the reconciliation bug class — is a
compile error for any typed server caller.
Approach (no build step): Node 25 runs .ts natively (type-stripping).
Migrated modules use the explicit-ESM .mts extension (the project is
"type":"commonjs", which disables .ts ESM auto-detection) and are required
from the CJS callers via Node's require(esm) — verified working. Infra:
tsconfig.server.json (allowJs + checkJs:false → incremental like the
client), npm run typecheck:server, check:server extended to .mts, wired
into ci. 28 require sites repointed to money.mts.
typecheck:server clean; full server suite 225 + e2e probe 17/17 green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The maintained line, and it forwards rejected async handlers to the error
middleware natively — closing the Express-4 gap where a throw in an async
route became an unhandled rejection that hung the request (this is why no
throwaway asyncHandler wrapper was needed). Small surface: only three
bare-* routes needed the path-to-regexp-v8 named-wildcard form
(/api/*splat, /legacy/*splat, /{*splat}); no removed APIs in use, all
req.query uses are reads. Suite 225 + probe 17/17 green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
As files convert to TypeScript they must keep the react-hooks enforcement.
Added a .ts/.tsx config block using the typescript-eslint parser with the same
rules-of-hooks/exhaustive-deps/react-refresh rules; swapped core no-undef +
no-unused-vars (type-blind) for TS-aware equivalents. Verified 'npm run lint'
traverses .ts and flags issues there; 0 errors on the existing .ts files.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adopting TypeScript incrementally (the move off plain JS). Upgraded jsconfig ->
tsconfig with strict + noUncheckedIndexedAccess, but allowJs + checkJs:false so
every existing .js/.jsx keeps resolving and running untouched — only .ts/.tsx
get type-checked. Added 'npm run typecheck' (tsc --noEmit) and wired it into
'npm run ci'. Installed typescript 6 + @types/react/react-dom 19. Client-scoped
(Vite resolves the '@/' paths); the CommonJS server is a separate TS story.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Added babel-plugin-react-compiler to the Vite React plugin. The compiler
auto-memoizes components/hooks at build time, so manual useMemo/useCallback
become largely unnecessary going forward (existing ones are left in place —
harmless, and the compiler adds the rest). Safe here because the codebase is
rules-of-hooks clean (enforced by eslint-plugin-react-hooks).
Validated: production build succeeds and the e2e probe passes 17/17 with the
compiler on — every authed page renders axe-clean and Tracker/Summary/Analytics
reconciliation holds. (Build time ~2.2s -> ~6.2s from the compiler pass; the
runtime win is automatic memoization.) The compiler ESLint rule needs
eslint-plugin-react-hooks v6 — a future upgrade.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
There was no linting at all — nothing enforced rules-of-hooks (conditional-hook
crashes) or exhaustive-deps (stale closures) across 125 client components/pages.
Added an ESLint flat config (eslint.config.mjs) scoped to client/, an 'npm run
lint' script, and the devDeps. First run: 0 rules-of-hooks violations (good),
6 errors + 13 exhaustive-deps warnings to work through next.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace all Save buttons on the Settings page with debounced auto-save:
- useAutoSave hook: debounce with latest-payload-wins, flush() for blur,
pending-edit flush on unmount, status machine (idle/saving/saved/error)
with saved fading back to idle. Covered by 6 Vitest tests (fake timers).
- SaveStatus pill (framer-motion) in the page header and notification card
headers — Saving…/Saved/Save failed.
- Timing per control: toggles/selects/channel ~150-400ms; typed inputs
(email, URLs, grace period, drift pct) 900ms + flush on blur.
- Push token never auto-saves mid-type: saves on blur only, so a partial
token can never overwrite a working one.
- Notification cards no longer refetch parent settings on save (would
clobber in-flight edits under auto-save).
- Decision: no undo toast — settings are non-destructive and instantly
re-editable; undo would add noise without safety.
- vitest include now picks up .jsx tests; jsdom + @testing-library/react
added as devDependencies.
- Extracted known-service catalog to dedicated /subscriptions/catalog route
- Simplified main Subscriptions page to focus on tracked services + bank-backed recommendations
- Replaced inline Pause/Resume with Edit + MoreHorizontal dropdown on subscription rows
- Added 'Improve Matching' card linking to Service Catalog
- Vite proxy respects API_PORT env var for dev flexibility
- Added top_200_us_subscriptions_researched dataset
- Updated HISTORY.md with v0.35.0 changes
null