jacobh2
ad0e3c0b33
A yearly-repeating period kept the same calendar date instead of the same
weekday: 2026-10-05..10-11 (Mon-Sun) recurred as 2027-10-05..10-11 (Tue-Mon)
rather than 2027-10-04..10-10. The snap added in 224fe7a only ever covered
window_mode == 'fixed'; 'flexible' wishes were unaligned at both stages.
project_window now snaps both modes. A flexible bound is still a bounding
range, but it keeps its weekday alignment year over year with its span held
fixed, so the search space never shrinks and a bound that is exactly as long
as the target (its only candidate) repeats week-shaped.
_rank_candidate gains a Monday-start term between cost and start, so equal-cost
candidates break on week alignment instead of "earliest". It sits after cost
deliberately: a cheaper non-Monday start still wins, because red days outrank
alignment.
_reuse_already_funded_window would otherwise freeze a pre-existing misaligned
placement forever, so it now releases a year whose reuse winner is misaligned
when a fresh search ranks strictly better -- never when one of its days is
pinned or excluded, so a deliberate per-day override always survives.
Migration 008 realigns flexible series already materialised under the old rule;
projection runs once, at create/update, so nothing repairs them on read.
Alongside this, the planner now says when it moved a period to catch red days.
WindowOutcome carries shift_gain_days (the saving over the naive bound-start
placement) and the calendar toasts it, once, on the sync that decided it.
A fixed window is never auto-moved -- its "exact requested placement,
unconditionally" guarantee stands -- but its shift suggestion is no longer
hidden behind a shortfall, so a fully-funded one can surface a cheaper week too.
Editing a period in the popover now also sends window_mode: 'fixed' with a
matching target. Preset chips create flexible wishes; dragging one down to an
exact week left it flexible with a week-long bound, which is the degenerate
single-candidate case that produced the reported dates.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DXPBfr1v7F71t6QcbHCaaT
Föräldraledigplannerare
A planner for Swedish parental leave (föräldraledighet) — solo or with a partner — that folds in vacation days and optimizes leave over time so that as few days as possible are lost at the child's 4th birthday, after which only 96 days may be saved (48 per parent, non-poolable, as soon as the child has two or more parents).
Full-stack monorepo: TypeScript/React/Vite frontend, Python/FastAPI backend,
SQLite by default. Postgres is a partial stub, not a usable swap yet — see
backend/README.md.
Quick start
From frontend/:
npm install
npm run dev # runs uvicorn (:8000) and Vite (:5173) together
Open http://localhost:5173 (not 127.0.0.1 — Vite's dev server binds
IPv6 ::1, so only localhost resolves to it).
The backend can also run on its own, from backend/:
uv sync
uv run uvicorn app.main:app --reload --port 8000
Seed a ready-to-explore demo workspace (couple, a partner, a child, vacation history, leave wishes, a deliberate shortfall) through the normal service/repository layers — no raw SQL:
cd backend
uv run python scripts/seed_demo.py --reset
It prints the demo login (fixed credentials, safe to re-run — without
--reset it's a no-op if the demo user already exists).
The Försäkringskassan scraper (BankID login flow via Playwright) is enabled
by default. Exercise the login UI directly with npm run login from
frontend/; Chromium must be installed separately:
uv run playwright install chromium.
The app
Three tabs:
- Översikt — the headline figure for days at risk at the 4th birthday, per-parent bars against each parent's savable cap, a year-by-year timeline, and shortfall/surplus detail below the fold.
- Kalender — where leave is authored. Drag a range to create a period, click a band to edit it, click a day to change which resource funds it (sjukpenning/lägstanivå/vacation). A season-preset rail helps place common periods (summer, Christmas, Easter). Day overrides (marking a specific date as a working day or a free day, overriding the calendar) live in Inställningar, not on the calendar itself.
- Inställningar.
The calendar is the plan. There's no separate propose-then-apply step:
every edit recomputes and persists immediately, reporting what changed with
an undo. A LeaveWish is a recurring intent (e.g. "summer", "Christmas");
the allocator decides which resource — parental days, vacation, unpaid —
funds each day, without retyping days a user has pinned by hand.
First run walks a wizard: add the child, optionally add a partner (planning solo vs. together is derived from whether a partner has been added — there is no separate solo/couple choice), then set balances — fetch from Försäkringskassan via BankID, enter manually, or skip (figures are then marked as estimates).
Layout
backend/app/
entities/ persistence-layer row shapes
models/ Pydantic v2 API schemas
repositories/ abstract repository interfaces
repositories_impl/ sqlite/ (default, complete) and postgres/ (partial — one entity only)
services/ business logic — planning, tips, analytics, auth
scrapers/ Försäkringskassan scraper (Playwright)
routers/ thin FastAPI route handlers
backend/migrations/ Alembic migrations (hand-written, not ORM-generated)
backend/scripts/ seed_demo.py and other dev-only scripts
frontend/src/
pages/ route-level screens (Dashboard, Calendar, Settings, ...)
components/ presentational + composite components (BEM CSS)
hooks/ data-fetching/state (the data layer — no TanStack Query)
services/ api.ts (HTTP client) + errors.ts
context/ Auth + Workspace React context
frontend/e2e/ Playwright end-to-end specs
Routers call services; services call the repository interfaces — never
repositories_impl directly. Components consume hooks; hooks call
services/api.ts — components never call fetch directly.
Tests & gates
Backend (from backend/):
uv run ruff check .
uv run ruff format --check .
uv run pytest --cov=app --cov-report=term-missing --cov-fail-under=80
Frontend (from frontend/):
npm run typecheck
npm run lint # ESLint, zero warnings allowed
npm run coverage # Vitest, thresholds: lines 96% / functions 91% / branches 92% / statements 96%
Both suites pass comfortably above their gates (backend ~84% vs. an 80%
floor; frontend above 96/91/92/96 — see frontend/vitest.config.ts). Gates
are a regression ratchet, not a target; run the commands above for current
numbers rather than trusting a hardcoded count here, which drifts fast.
CI (.github/workflows/ci.yml) runs all of the above, plus a Playwright E2E
suite that reports but does not yet block merges.
Further reading
CLAUDE.md— conventions, stack details, model-tier rules for contributors.ARCHITECTURE.md/PLANNING_ENGINE_SPEC.md— design and rationale. Both predate parts of the current behavior; sections marked SUPERSEDED note what changed and why.backend/README.md— backend-specific setup and the reference vertical slice to copy when adding a new entity.