Один чек часто смешивает несколько категорий трат (гипермаркет: часть — продукты,
часть — бытовая химия). Раньше такую транзакцию приходилось огрублять до одной
категории или дробить вручную на несколько записей. Теперь у транзакции есть
второй режим: 2+ доли (TransactionSplit) с суммами по разным категориям, которые
обязаны в точности складываться в сумму транзакции.
- Новая модель TransactionSplit + миграция.
- split_rule_violation() проверяет инварианты на создании и редактировании.
- category_rollup.py — общий helper, которым теперь пользуются Dashboard, отчёт
по категориям и бюджеты, учитывая доли сплитов наравне с обычными транзакциями.
- Фильтр по категории в списке транзакций находит и сплит-транзакции.
- Сплиты попадают в полный бэкап/восстановление БД.
- 13 новых backend-тестов.
- Фронтенд: переключатель «Разбить на несколько категорий» в форме транзакции
с повторяемыми строками и индикатором остатка суммы; строка транзакции в
списке показывает все категории сплита.
`AURUM_CORS_ORIGINS` defaults to `*`, and the middleware was configured with
`allow_credentials=True`. Starlette does not answer that combination with a
literal `*` — it echoes back whatever `Origin` the caller sent and adds
`Access-Control-Allow-Credentials: true`, so any page open in the same
browser can call the instance, with the user's Basic Auth attached, and read
the response:
GET /api/accounts Origin: https://evil.example
-> Access-Control-Allow-Origin: https://evil.example
Access-Control-Allow-Credentials: true
That undercuts the README's advice that leaving Basic Auth off is fine while
the instance is only reachable from localhost: reachable-from-localhost
includes every website the user visits.
Nothing in the shipped setup is cross-origin. Compose serves the UI and the
API from the same nginx, and the Vite dev server proxies `/api`
(frontend/vite.config.ts), so neither deployment nor development needs CORS
at all. The default is therefore empty — no cross-origin access — and the
middleware is only installed when origins are configured deliberately.
An explicit `*` is still honoured for a genuinely public read-only setup, but
never together with credentials, which is what the spec intends `*` to mean
anyway.
get_budget_status() summed only transactions whose category_id matched the
budgeted category exactly, while the Dashboard breakdown and the category
report both roll a subcategory's spending up into its parent
(coalesce(parent_id, id) in dashboard_service.py and reports_service.py).
With one level of subcategories — the app supports exactly one — a budget on
"Housing & Utilities" ignored everything filed under "Rent": the Dashboard
showed 900 spent for the month and the Budget page showed 0 of 1000 for the
same category, same month. Since insights_service's over-budget alert reads
this same status, the alert stayed silent too.
Budgets now count the budgeted category plus its children, matching the
rollup the other two screens already use. A subcategory that has a budget of
its own keeps tracking its own spending, so both bars move against their own
limits — that case is pinned by a test as well.
TransactionRead inherited TransactionBase's model_validator, so the rules a
new row must satisfy were also applied on the way out, to rows already in
the database. A row can stop satisfying them through no fault of its own:
transactions.transfer_account_id is ON DELETE SET NULL, so deleting an
account leaves every transfer that pointed at it without a destination.
The result was that deleting an account could 500 the entire transactions
list — permanently, and with the list gone there was no way in the UI to
find and delete the offending row.
The fields move to TransactionFields, which reading uses; TransactionBase
keeps the invariants for the write path. The frontend already tolerates a
missing destination (TransactionsTable renders the transfer suffix only when
transfer_account_id is set, and the edit modal makes you pick a destination
before saving), so such a row is now visible, editable and deletable
instead of fatal.
Matches CoinMarketCap's flow: clicking anywhere on a coin's row in the
holdings table opens its transaction history (previously only a dedicated
icon button did). The table's own History icon is gone since the row now
does that job; the remaining "+"/delete icons get stopPropagation so they
don't also trigger the row click.
Adds the missing piece: transactions could be viewed and deleted but not
edited. New PATCH /crypto/transactions/{id} (partial update, never calls
CoinGecko, same oversell guard as creating a sell) backs a pencil button on
each history row that opens CryptoTransactionModal pre-filled — the same
dialog used for "+", now doing double duty as create/edit.
86/86 backend tests passing (4 new). Verified live against the real API:
create, edit quantity/note, 404 on unknown id, cleanup — real user data
untouched throughout.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01855PiW5Wd8n4aRW4n9kByY
Adds the analytics half of the CoinMarketCap reference: a value-over-time
chart (GET /crypto/history, forward-filled from AssetValuation like Net
Worth's own series but scoped to crypto assets), an allocation breakdown by
coin, and stat tiles for all-time profit, total invested (cost basis), and
best/worst performer.
Deliberately no "24h" range — history is only as dense as the sync cadence
(button or once/day), so a 24h chart would be one or two points, not a
smooth intraday line. Cost basis / all-time profit / best-worst performer
aren't separate endpoints: they're trivial to derive client-side from
fields already in GET /crypto/holdings, so no redundant backend logic.
Allocation card follows the dataviz skill's categorical rule: top 7 coins
by value get a fixed-order color from the app's already-validated 8-slot
ramp, an 8th+ folds into "Other" rather than generating a new hue.
82/82 backend tests passing (3 new for /crypto/history). Verified live
against the real CoinGecko API — and caught a real holding the user had
already added directly through the UI between messages, which the smoke
test correctly left untouched while cleaning up only its own test data.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01855PiW5Wd8n4aRW4n9kByY
Replaces the plain quantity field with a CryptoTransaction log (same
"record events, derive the total" shape as Goal/GoalContribution): quantity
and average buy price are now computed via weighted-average cost — a buy
blends into the running average, a sell reduces quantity but leaves the
average cost of what's still held unchanged.
Switches price fetching from /simple/price to CoinGecko's /coins/markets
(still one batched call for every coin), which also returns 1h/24h/7d %
change in the same request. Old PATCH /crypto/holdings/{id} is replaced by
POST .../transactions (buy/sell, rejects selling more than held), GET
.../transactions (history), and DELETE /crypto/transactions/{id}.
Frontend: CryptoHoldingsTable replaces the simple list with a real table
(Name/Price/1h%/24h%/7d%/Holdings/Avg Buy Price/Profit-Loss/Actions),
horizontally scrollable on mobile rather than dropping columns. New
CryptoTransactionModal (buy/sell) and CryptoTransactionHistoryModal.
Backend: 79/79 tests passing (13 in test_crypto.py, up from 9). Also
verified live end-to-end against the real CoinGecko API with the user's own
key — create, buy, sell, oversell rejection, history, manual refresh, all
against real BTC/ETH prices in EUR — test data cleaned up after.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01855PiW5Wd8n4aRW4n9kByY
Track individual coins (identity + quantity) in a dedicated tab, backed by
CoinGecko's Demo API (AURUM_COINGECKO_API_KEY). A holding is just another
Asset under the hood (asset_class=crypto) with its value written as a
regular AssetValuation — the existing Net Worth aggregation needs zero
changes to pick it up.
Prices never refresh in the background (no scheduler in this stack, same
reasoning as recurring transactions): a "Refresh prices" button forces a
sync, and GET /crypto/holdings lazily re-syncs at most once every 24h.
CoinGecko outages leave existing values untouched (error_key: "unreachable")
instead of failing the request. Backup/restore now carries CryptoHolding
rows too, so a restore doesn't silently strip a holding's coin identity.
Backend: 75/75 tests passing (9 new), including a real identity-map
staleness bug the tests caught in the raw-Core-upsert + ORM re-read pattern,
fixed via populate_existing=True.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01855PiW5Wd8n4aRW4n9kByY
`GoalContributionCreate.amount` used `Field(ne=0)`, but pydantic has no
"not equal" constraint — the keyword was accepted as unknown schema
metadata and never validated anything. `POST /api/goals/{id}/contributions`
with `"amount": "0"` returned 201 and wrote a contribution that moves the
goal's progress by nothing, and the comment right above the field says zero
is pointless.
Replaced with a field_validator so the rule is enforced, deliberately not
with `gt=0`: negative contributions are intentional (money taken back out of
a goal), and the added tests pin both halves of that.
This also clears the `PydanticDeprecatedSince20: Using extra keyword
arguments on Field (Extra keys: 'ne')` warning that the test suite printed
on every run — that warning was the symptom.
Claude-Session: https://claude.ai/code/session_01UmuMDFrwdvQkce885VaaTh
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
The transfer rules — destination required, destination different from the
source account, no category, and no destination on a non-transfer — were
enforced by TransactionBase's validator on create, but the PATCH route only
re-checked category-vs-type.
So `PATCH /api/transactions/{id} {"type": "transfer"}` on an ordinary
expense committed a transfer row with transfer_account_id = NULL, and only
then failed serializing the response (TransactionRead inherits the same
validator). The caller got a 500 and would reasonably assume nothing was
saved — but from that point on every `GET /api/transactions` failed the same
validation, so the transactions list, the app's main screen, returned 500
permanently, with no way left in the UI to delete the offending row.
The rules now live in one place, transfer_rule_violation(), used by both the
create schema and the update route, so the two paths can't drift apart
again. The update path validates the row as it would look after the patch
rather than only the fields sent, since switching type alone invalidates
fields that were left untouched.
The frontend already sends a complete payload on edit (explicit nulls for
category_id / transfer_account_id) and does these same checks client-side,
so this only closes the hole for direct API calls — no UI change needed.
Claude-Session: https://claude.ai/code/session_01UmuMDFrwdvQkce885VaaTh
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Fourth of the four ideas picked from the Maybe/Sure feature audit, and the
last one queued up. A 3-step wizard (account -> column mapping -> preview)
at /transactions/import, parsed entirely client-side (new lib/csv.ts —
delimiter auto-detect, 5 date formats, thousands/decimal-separator-aware
amounts) and submitted through a new all-or-nothing POST /transactions/bulk
endpoint, mirroring backup restore's validate-everything-before-touching-
anything approach.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Third of the four ideas picked from the Maybe/Sure feature audit. Tags are
orthogonal to categories (category = what kind of spend, tag = which
event/project) — many-to-many, created inline while typing in the
transaction form, filterable from the Transactions page.
Also adds tests/test_backup.py, the first test coverage the backup
export/import path has ever had, and fixes a MissingGreenlet bug it
surfaced in restore_backup's tag-wiring.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Two of the four ideas picked from the Maybe/Sure feature audit:
- Idle cash: a 5th proactive alert flags a depository account (checking/
savings/cash) sitting at or above a configurable balance with no
transaction touching it for a configurable number of days.
- Category hierarchy: categories can now have one level of subcategories
(Groceries -> Alcohol). Dashboard/report aggregations roll a
subcategory's spending into its parent's slice.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Fresh installs (this app is published on GitHub) only got Salary,
Freelance, Investments, Gifts, Other Income by default — adds Business
Income, Rental Income, Benefits, and Item Sales, rounding out the 8-slot
categorical palette. Only affects new databases (seeding is a one-time
no-op once any category exists).
Also fixes CategoryList.tsx (added last commit) to run category names
through translateCategoryName() like every other category display in the
app, and registers the new defaults' translation keys.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Transactions/recurring templates entered with mixed casing ("траты на
продукты") are now capitalized on create/update via a Pydantic validator,
so the fix applies to every client, not just the web UI. Adds a one-off
backfill script for existing rows.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Lets the user find a past purchase by keyword instead of paging through
months manually — search hits description/merchant/notes and ignores the
month filter, with a "jump to month" action on each result to view it in
its normal monthly context.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Asset(...) в create_asset игнорировал payload.risk_level, а _to_read()
не передавал risk_level в AssetRead — из-за дефолта в схеме API всегда
отвечал Medium вне зависимости от выбора в форме и данных в БД.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Transactions can carry a `notes` field (set via the API — e.g. the
currency-conversion metadata on this user's imported historical data), but
there was no way to see it in the UI at all.
TransactionsTable now shows a note icon next to edit/delete, only when a
transaction actually has one, opening a dialog with the text. The edit
form's `notes` field was already wired end to end (state, payload) but had
no input in the JSX — added it, since a view-only note with no way to edit
or clear it would be a dead end.
Also fixed e2e/run.sh's readiness check, found while running the full
suite for this change: it polled `/` (nginx, answers almost immediately)
instead of `/api/health`, so the first test could start before the backend
finished migrating/seeding and fail trying to parse nginx's HTML error
page as JSON.
Picking a category at the top meant scrolling past the whole "top
categories" list before reaching its chart, then scrolling back up to
pick another one. Swapped the two so the chart renders immediately after
the category/range selector.
Language and Theme were two separate Cards stacked on top of each other —
both are pure client-side display preferences, so splitting them side by
side (language left, theme right) saves a card's worth of vertical space.
Collapses back to a single column on mobile.
Theme previously followed only the OS's prefers-color-scheme, which
switches automatically on this user's Mac at sunset/sunrise. Adds an
explicit override in Settings: Light/Dark/System, System being the
previous (default) auto-follow-the-OS behavior.
Persisted client-side in localStorage (lib/theme.ts, same pattern as
language) rather than server-side like currency — it's a pure display
preference, and synchronous localStorage reads are what make the
anti-FOUC inline script in index.html possible. index.css already had the
[data-theme] CSS variables laid in ahead of time; only the JS to actually
set the attribute, and the Settings UI, were missing.
Covered by 2 new Playwright specs (theme-switcher.spec.ts) run against the
isolated e2e stack: explicit choice overrides a dark OS and survives
reload; switching back to "system" doesn't leave a stale forced theme.
backend/tests/: 33 pytest tests against a real Postgres database
(aurum_test, created/migrated/dropped per run) exercised through the actual
HTTP API — not sqlite, not mocks, and the real `aurum` database is never
opened (app lifespan is skipped, get_session is overridden per test).
Covers transaction category-kind validation, year/month/date-range
filtering and pagination, dashboard summary arithmetic, budget status, cash
flow, and the /transactions/years picker.
e2e/: 6 Playwright specs driving a real browser against a fully separate
throwaway docker-compose stack (run.sh: separate project name, database,
port, and volume, torn down with `down -v` on exit) — regression coverage
for the Recent Transactions month bug, the Card grid-overflow bug, the
frozen-year bug, the Net Worth default range, and the Dashboard->
Transactions deep link, all fixed earlier this session.
Also synced the backend's APP_VERSION constant (stale at 0.15.2, embedded
in exported backups) with the frontend version.
Year was previously hardcoded to the current year on all three pages
(useState with no setter) — only the month could be changed via
MonthSelector, making it impossible to look at or compare older years
even though transaction history goes back further than the current
year.
Adds GET /transactions/years (contiguous range from the earliest
transaction through the current year) and a YearSelector component
matching MonthSelector's style, wired into all three pages alongside
a real year useState.
Aurum — a self-hosted personal finance operating system: cash flow,
net worth tracking, budgets, goals, recurring payments, reports,
proactive alerts, and a compound-interest returns calculator, all
bilingual (RU/EN) and mobile-first. Ships as a three-container Docker
Compose stack (Postgres + FastAPI + nginx-served React frontend).
See README.md for setup and CONTRIBUTING.md for development.
