The count moved three times on 2026-08-12 and the history is the point:
755 when the predicate was written
904 after board 219 turned unstated zeros into NULLs — those 149 rows had
been failing the "no amount" test on a technicality
699 after 210a retired 205 Amazon "share your experience" mails, which were
never lifecycle events at all
620 after 210b joined 79 phantoms to their real parent, placebo-verified at
a 1.9% error rate
What is left is the corpus coverage ceiling rather than a defect — a mail whose
parent is not in the corpus has nothing to join to — and the tooltip now says
that instead of describing the whole class as an open ingestion defect.
Phase 1 read linkage out of expense_metadata, which is shaped as ONE ROW PER
TRANSACTION — transaction_id UNIQUE, matched_transaction_id partial-unique — so
every multiplicity it expresses is smuggled through a string key (0029 keys
split shipments <entity_key>#f<fact_id>). A BNPL plan needs four rows against
one order and has no such trick available.
Migration 0030 adds order_transaction_links (many-to-many, keyed on entity_key
rather than entities.id: finance-app does not model the spine and must not hold
an FK across a boundary a re-extraction can decompile) and backfills all 63
existing bridge rows.
THE UNIQUENESS RULE WAS WRONG FIRST TIME, in the most on-brand way available.
It read (entity_key, leg_kind, COALESCE(leg_index, 0)), so two shipment legs of
the same order — both leg_index NULL — collapsed to one key and ON CONFLICT DO
NOTHING dropped one SILENTLY. Caught only because the backfill reported 62
against 63 candidates. The row it ate was order_amazon_249-4859367-0690246's
$130.00 second shipment, the same order named in migration 0029's comment as
the reason split shipments need distinct keys at all. What identifies a leg
depends on its kind: an instalment by its INDEX, a shipment by its FACT, a
whole-order charge by neither.
ALL FOUR READERS MOVE TOGETHER, links first with expense_metadata as fallback:
- LINK_LATERAL (list txn_count/first_txn_id, and the has_transaction facets
that read it)
- the detail page's transactions query
- /api/transactions/[id]/order
- the order_ctx lateral in queries.ts
Moving fewer is not a smaller change, it is an inconsistent one: the matcher's
four links for order_ebay_14-11714-95953 come with no expense_metadata row, so
a half-move would show four instalments on the detail page while the list said
txn_count = 0 and put the order on the wrong side of BOTH has_transaction
filters. Verified after: detail 4 legs, list txn_count 4, has_transaction=yes
includes it, =no excludes it.
UNION not UNION ALL on transaction_id — after the backfill the same charge is
legitimately in both stores and counting it twice would show "2 charges" on a
single-payment order.
order_platform is deliberately NOT coalesced with the link's platform. It gates
the receipt disclosure arrow on /transactions, and a BNPL leg has no receipt
behind it — filling it in put an arrow on four Afterpay rows that expand to
nothing, which is the exact promise the arrow exists to avoid making. Separate
leg fields carry the sub-line instead ("1 of 4 - DJI Air 3 Fly More Combo").
order-details.tsx now also requires a real receipt (platform present) before
rendering, because the endpoint can answer with a link alone.
Two invariant views, both empty and expected to stay so: order_link_orphans
(spine re-keying silently orphans a TEXT key — bridge links are rebuildable,
`manual` ones are lost curation) and order_link_drift (the two stores
disagreeing). The plan's suggested fix for drift — widening the bridge's NOT
EXISTS guard to "neither store has it" — was NOT taken: the expense_metadata
INSERT has no ON CONFLICT, so re-attempting an order that already has a receipt
row would duplicate it. Detecting is cheap; a non-idempotent re-write is not.
Unchanged: order_feed 6,265, order_spend AUD 4,895 / $442,651.80. Links 75.
Migration 033 adds entity_orders.summarises_period; order_spend drops those
rows and order_feed keeps them. Unbadged, a $2,376 annual tax receipt
restating twelve monthly donations renders as the largest single purchase of
the year, so both surfaces say what it is:
- /orders — a "year summary" chip beside the auth badges
- /orders/<key> — a banner in the same shape as the settled-rail one, saying
the monthly debits are in the ledger under their own dates
The amount is deliberately still shown. It is the year's giving total and the
reason the row is worth keeping; it just belongs to no total on the page. The
detail page already falls back to o.order_total when order_spend has no row,
which is exactly the case now — verified: gross_total null, order_total 2125.5.
Board 219 nulled 263 unstated zero totals in the spine, which moved 149 rows
into the lifecycle-only hide predicate — they had been failing its "no amount"
test on a technicality, an unstated 0 they should never have carried, while
matching the phantom signature on every other count.
Tried exempting rows that carry line items, on the theory that naming what was
bought proves a purchase. It surfaced 442 rows and the first window checked
showed why it is wrong: the Amazon "Ordered:" mail that mints a msg- phantom
carries the item list too, so Xelsluthe and YIWENTEC each rendered TWICE again
— the exact complaint this page was fixed for today. Reverted, and the reason
recorded in place so it is not re-attempted.
Items do not separate an orphan purchase from an item-carrying echo. Only
deduplication does, and that is board 210.
No behaviour change from the previous commit; comments and counts only.
order_merchant_cadence.order_count is a count(*), so BIGINT, and JSON.stringify
throws on it: 'Do not know how to serialize a BigInt', a 500 carrying no SQL.
It fires only where a cadence EXISTS, so the failure hid behind the
merchant_entity_id fault and reappeared the moment that was fixed — St. Ali
(40 orders, 14-day cadence) still 500'd while Amazon opened fine.
getOrderDetail joined order_merchant_cadence on merchant_entity_id. Migration
021 rekeyed that view to merchant_key (merchant_entity_id is NULL on ~9% of the
feed, and duplicate merchant entities split one shop's history); 027 and 029
moved order_feed and order_spend across, and this query was missed. Every
/orders/<key> request has since failed with
column rec.merchant_entity_id does not exist
The reason nobody saw a 500 is the second half of this commit: the page
collapsed every failure into 'That order could not be found.' A server fault
wearing the costume of a data condition reads as an empty spine and gets
investigated in the wrong repo. The hook now carries the status and only a
genuine 404 says the order is missing.
source_trust is 'untrusted_external' on 100% of rows — every order came from
email — so the badge marked every row and discriminated nothing. auth_verdict
(migration 029) does: 93% pass, and the 7% that do not are the ones worth
seeing. Adds the injection-scanner flag beside it.
Rebuilding the prototype in React quietly dropped most of what made it read
well. Side by side, the live page had lost the year strip entirely — which was
both the hero and the context for the date range — along with the masthead
totals, the A$ prefix on amounts, and the Fraunces wordmark. The trust marker
rendered as a tofu box. Restored, with the strip clickable so a bar scopes to
that year.
Also three content fixes, all visible the moment anyone looked at a real page.
The title and the manifest were printing the same text twice on 1,964 of 4,649
itemised rows — 42% — because on a single-item order the order title IS the line
item. The title now earns its line only when it says something the items do not.
Food line items carry the full customisation: one Subway order runs to 450
characters naming every topping, and it swamped everything around it. Rows show
a short form, cut at the first bracket where there is one, since "Footlong
(Italian Herb…" is the product and the bracket is the customisation.
Rows now expand, following the disclosure pattern the transactions page already
uses, and the query fetches twelve items rather than three so opening one costs
no round trip.
Merchant names get a presentational tidy — corporate suffixes dropped, a
lower-cased domain capitalised. This never merges two merchants: the estate
holds amazon.com.au, Amazon.in and Amazon Services Australia as three distinct
entities, and Amazon.in must stay separate because it is a different marketplace
rather than a name variant. Actually unifying them is the merchant-alias bridge.
The first cut rendered entity_orders and nothing else, and it read like a table
dump: a courier tracking notice and a subscription payment both presented as
retail purchases, merchants shown as sender-domain slugs, and no indication that
the same coffee roaster had been billing fortnightly for two years.
Rows now carry what the ingestion layer already knew. A kind badge from
content_class distinguishes a delivery notice, an invoice, a booking and a
subscription from an actual purchase — NULL stays unbadged rather than being
labelled a purchase, because 48% of rows predate the interpretation index and
"unknown" is not "order". display_name shows the resolved merchant (98% of rows)
instead of the slug. A recurrence badge marks merchants billing on a cadence,
which is derived from the gaps between orders rather than stated anywhere in the
mail. And the order's own title now appears when it differs from the merchant,
so the manifest is not the only detail on the row.
Adds the services lane for bookings, invoices and subscriptions — things with no
goods and no delivery, where the retail columns are dead space.
The spine holds ~6,300 purchase orders back to 2006, ~4,600 of them itemised,
and 61 reach a transaction. Everything else has been visible only through SQL.
This adds /orders and /orders/[entityKey] over it.
The point of the page is the manifest. /transactions can only ever say "AMAZON
AU MARKETPLACE SYDNEY"; a row here says what was in the box, which is the one
thing the ledger structurally cannot carry.
Five lanes, because the shapes genuinely differ — retail ends refunded or
returned 14.1% of the time against food's 5.1%, food has no meaningful ETA where
grocery has one on 54.8% of orders, digital never ships at all. The lane comes
from order_lane() in migration 018 rather than a column, because slug 'uber'
carries 394 taxi rides and 485 Eats orders.
Defaults to this year: 449 orders rather than 6,283. Twenty-one years is the
archive, not the working set.
Three things the data forced. Unknown amounts render "not stated", never $0.00,
because 1,648 orders have no amount and a zero would be false. A full reversal
strikes the figure through; a partial refund does not, since striking $191.40
when $13.33 came back is a lie — the charge stays primary and the credit sits
under it with the net. And rows with no amount, no reference and one lifecycle
event are hidden by default, which lifts amount coverage from 74% to 84%; the
toggle says on its face that it is a workaround for board 210 rather than a fix.
Reads are raw SQL in lib/order-feed.ts rather than queries.ts: the spine is
written by the ingestion-engine, is not in prisma/schema.prisma and never will
be, and mixing it into a file where everything is Prisma-modelled would destroy
that invariant.
/orders is gated by an explicit viewer allowlist. Not because the three people
listed need protecting from each other — everything here is on one person's
cards — but because a participant is an accounting entity and any participant
row with an email is a login. Adding someone to split a holiday must not
silently hand them the purchase history.
Verified live: gate returns 403 for non-participants and for a missing identity
header; search "drone" finds the DJI order through its line items; the detail
page renders its three lifecycle events and its Afterpay settlement sibling; and
a bridged Amazon order shows both split-shipment charges.
The table rendered every split row (1,279 today, ~600/yr growth), so page
weight was unbounded even after the re-render fix. Render the first 100 of
the filtered/sorted set with Show more (+100) / Show all; any filter or
sort change resets the window. Data stays fully client-side on purpose —
search still counts matches across all history — only the DOM is capped.
Headless-measured expand stall: 165ms -> 35ms.
Expanding a receipt on /shared froze the browser: this table renders every
split row at once (1,279 today), and one state change re-rendered all of
them. 520ms of each pass was formatDate constructing a fresh
Intl.DateTimeFormat per call (2,558 calls per render); the rest was
rebuilding 1,279 rows to change one. Headless-measured click stall:
802ms -> 165ms on the server; slower machines multiply the former.
- module-level DATE_FMT, reused
- rows extracted into memoized SharedTxRow with stable callbacks
- filter+sort wrapped in useMemo so unrelated renders skip it
The item list existed only on the owner's transactions page — the person
a meal or order was split WITH could see the amount but never what was in
it. getSharedTransactions now carries the same order_platform LATERAL the
transactions page uses, and rows with a receipt get the same disclosure
arrow + OrderDetails expansion. The shared viewer is a split participant,
so canAccessTransactions already authorises the order API for them; meal
rows keep the verdict UI (the partner records their own), bridge rows
stay review-free.
Verdicts are the meal lane's feedback loop (loved/never steers the next
restaurant order and the nudge card warns on re-order). Bridge rows
(source='order-bridge': Amazon, eBay, Officeworks...) have no such loop,
so ItemVerdictToggle and OrderVerdict stay hidden there. The order API
now returns 'source'; OrderReceipt.platform widened to string — the
bridge writes real platform slugs beyond the meal trio.
Split-shipment orders are charged per shipment; the spine bridge now
writes one row per shipment sharing (source, order_reference). Meal-lane
idempotency semantics unchanged (index scope excludes only
source='order-bridge'); bridge rows get their own unique index on
(source, source_message_id). Applied 2026-08-10.
Spine-bridged receipts (Amazon shipment notices, eBay) name the goods
without per-item prices — null means 'the mail didn't say', and printing
$0.00 would assert it did.
The phase queries apply NET_SPEND_ROWS and EXCLUDE_RECONCILED_SOURCE like every
other analytic, so committed is $21,229.56 rather than the $22,050.51 a raw sum
gives. The $820.95 gap is a partial refund on the Luxury Escapes booking, and it
is netted out on purpose.
travel dominated every trip page and said nothing. The tempting fix is a finer
travel taxonomy, which needs a hand-maintained merchant list — the trap #19
already describes — and it is also the wrong diagnosis.
travel is the only category that spans both phases of a trip. Every other one is
100% on-the-ground: on Europe 2026, dining, transport, entertainment, groceries
and shopping are all exactly $0.00 before departure. The chart was not bad, it
was two economies stacked into one, and travel was the only thing visible in the
union.
So split on start_date and use the axis that carries information in each phase.
Booked ahead ($22,050.51, 57%) is all flights and stays, so merchant is the axis
— Agoda $4,490, Air India $3,454, Luxury Escapes $3,284. On the ground
($16,946.94) travel falls to $8,241 among dining $4,452 and transport $2,938, and
category is finally worth charting.
The hero is the ratio, not a lone total, with the on-ground daily rate beside it
— the only figure comparable between trips, since totals are not: Europe
$677.88/day against Auckland $83.39. A trip with near-zero committed spend says
so, because Sonu + Sunny's $184.84 is a filing artefact (both legs' bookings sit
on the first trip), not a cheap trip.
Two dataviz rules this page was breaking. Category bars now use one copper hue
with the name as a direct label: the per-bar rainbow double-encoded identity the
label already carries, and the trip subset fails CVD validation on this surface
(other vs shopping at delta-E 5.0 protan, below the floor of 6). And the hero
figure drops the serif and tabular-nums, which read as decoration at that size.
The phase bar is two ordinal steps of one hue, validated with --ordinal against
the card surface, with a 2px gap so the boundary is an edge.
278 passing, build clean. Data verified against the database directly; I could
not render the page in a browser to eyeball the layout.
You were right and my label was wrong. A payment is allocated to a trip as a
lump sum, and the grouped-payment allocation gave each trip enough to clear the
payer's gross share. So netting the other side off leaves a fully-paid trip
negative by exactly what the payment over-covered: Europe reads -$802.75 because
Sonu paid $8,004.04 against a net share of $7,201.30. That surplus is already
carried in the overall balance, where she still owes $5,313.38 — nothing is owed
to her, and "you owe them" said the opposite.
The arithmetic never changed; only what the page claims it means. paid_to_me is
the discriminator: negative with a payment into the scope is over-coverage,
negative with no payment is genuinely owed because the viewer's share of the
other person's spending exceeds theirs. Both cases now tested. All three of
today's negatives are the first kind.
The trip table gains an Overall balance column from the unscoped participant
balances, because a single trip's figure cannot tell you whether to pay anyone,
and the footnote now says settlement is always against the overall figure.
278 passing, build clean.
Shared view: the query already returned owner_name and effective_category, the
table just never rendered them. Paid by sits next to Splits because together
they are whose money went out and whose share it was. Search is client-side —
this endpoint returns all 1,267 split rows in one request with no pagination, so
there is nothing for a round-trip to narrow, and the sort was already
client-side. It matches description, merchant, notes, category and payer, but not
participant names: the dropdown does that, and "sonu" matching every row she is
split on would read as broken.
Trip owed collapses to one settle-up figure per person, with the breakdown
beside it so the net is auditable rather than asserted.
I argued against netting a few hours ago and was wrong. The claim was that the
grouped-payment allocation cleared each trip against the one-directional gross,
so netting would redefine that debt after the fact. The rows say otherwise:
Europe's $802.75 is 56 transactions Sonu actually paid across Rome, Venice, the
Dolomites, Bellagio, Lucerne and Paris on which I hold 25%, and paid_by_me is
$0.00 on every row of every trip because nothing has ever been recorded going
from me to her. Her side looked settled only because the allocation derived her
payment split from her gross, so it lands on zero by construction. The
one-directional view was hiding a live obligation, not protecting an allocation.
Nets now: Auckland Sonu +$1,077.25, Europe Sonu -$802.75, Sonu + Sunny -$936.34,
Europe Molina -$816.16.
Also correcting an error in my own reporting: I said Auckland's mirror was
$0.00. It is $428.39 — 17 Auckland rows Sonu paid that I hold a split on. Two
ad-hoc verification queries mis-joined on a nullable scope column and
under-reported the mirror side. The app code was never affected and the owed
column is still byte-identical.
The footnote now states the trap the netting exposes: a debt settled by a payment
left on the household tab still reads as outstanding on the trip. Payment 5
(Molina to Sonu, $1,605.49) is exactly that case and is left alone as a data
decision.
277 passing, build clean.
Trips were scoped to trips.owner_id, so Sonu saw no trips at all — despite
having paid for 104 of the tagged rows herself. Her own spending was invisible
on the only page organised around it.
A participant is now anyone with a split on, who paid for, or whose payment is
scoped to, a transaction tagged to the trip. Derived, not stored. A
trip_participants table was designed and rejected: the expenses already carry
the fact, and two records of one fact drift apart. Deriving it also excludes
Singapore + Bangkok 2026 from Sonu for free, which a table would have to be kept
in sync to do. Siddharth 4 trips, Sonu 3, Molina 1.
Everything about a trip is shared except delete. Both trip foreign keys are
ON DELETE SET NULL, so deleting Europe 2026 untags 210 transactions and NULLs
the trip scope on 6 payments — where the hand-derived Europe-first allocation
lives, which nothing recomputes. That stays with the owner.
Trip owed now returns both directions and nets neither. An obligation lives on a
row someone else paid for, so a viewer-as-payer figure can never hold it, and
Sonu's Europe read "you are owed $2,408.24" while omitting the $8,004.04 she
owed. Collapsing the two into a signed net is the tempting next step and would
have corrupted the scope allocation: the grouped-payment allocation cleared each
trip against the one-directional gross, so redefining the debt afterwards turns
$8,004.04 already allocated into an $802.75 over-allocation with household
understated by the same amount. Verified byte-identical — Auckland $1,505.64,
Europe Molina -$816.16, Europe Sonu $0.00, Sonu + Sunny $0.00.
getTransactions gained trip_all_rows so a participant sees the whole trip. It is
opt-in and not implied by trip_id, because the same endpoint backs the main
transactions list and its trip filter must keep owner scoping. Participation is
re-checked in SQL, so passing the flag for someone else's trip returns nothing.
Payments can finally say what they settle. trip_id has existed since migration
0022 but POST never read it and GET never returned it, so every payment made in
the app landed on household and the 9 trip-scoped rows were hand-written SQL.
"Both" needs no new shape — one row per scope sharing a linked_transaction_id.
Three write paths had no authorisation at all and were reachable by any
participant: assignTransactionsToTrip checked nothing, DELETE on a payment
deleted by bare id, and POST accepted any from/to pair. All three now check.
Also fixes the test suite, which was pointing at postgres-pantry: container IPs
move on recreation and 172.22.0.47 stopped being postgres-personal. It only
failed safe because the credentials did not match — resetDB now refuses to
truncate anything not named personal_test.
22 new tests, 276 passing, build clean.
The review was written 2026-07-26 and last touched 2026-07-29, one day before
the board replaced ACTIONS.md, so it was never migrated. Priorities 1-4 existed
only in this file. They are now epic #154 with #155-#159 under it.
Two claims in the doc were wrong. It said Priority 0 was guarded by
analytics-sql.test.ts; that file covers the three SQL fragments and nothing
else, so items 2, 3 and 6 -- pace/headline parity, the fees date range, and
partial-month comparison -- have no test at all. Those are the three where a
regression is silent, which is why the gap is worth a ticket rather than a note.
And it predated two changes in the same metric-integrity family: the signed
investments line (c70d2b1) and transfers hidden by default (f6c500b). The first
matters to the doc directly -- the cashflow strip it describes now carries a
signed invested figure, so a negative month is data rather than a bug.
The CSV importer has no idempotency and structurally cannot have any:
batchInsertCSVTransactions assigns row_index = MAX(row_index) + 1, and
row_index is the fifth column of uq_transaction_identity, so the constraint
is guaranteed a fresh value on every run. The same file imported twice
produces two full sets of rows and nothing objects.
Tolerable for the hand-driven bank CSV this was built for. Not tolerable for
any recurring feed, whose windows overlap by design — and not cleanable
afterwards, since every child of transactions is ON DELETE CASCADE, so a
duplicate must be superseded rather than deleted (0023). ~$42k of re-imported
statement rows already show how that goes.
source_ref carries the provider's own key verbatim, enforced by a partial
unique index rather than an application-side ON CONFLICT that a refactor can
quietly drop.
Found while evaluating Frollo (DECISIONS.md ING-11 in the smarthome repo).
That feed is not being ingested — 88% of it duplicates existing statement
coverage — but this gap is real independently of it.
A withdrawal from a fund is a disinvestment, not income: units convert
back to cash and net worth is unchanged. Summed unsigned it read as more
money invested. March 2026 showed $38,615.34 of investing in a month that
was net -$11,384.66, because a $25,000 Raiz withdrawal was added to an
$8,563.80 IBKR deposit instead of cancelling it.
Each credit costs twice — once for being added, once for not being
subtracted — so the error is double the credit: $50,000 in March, $3,000
in May 2025, $53,000 across the window. Since net = income - spent -
investments, March's net of -$55,918.73 should read -$5,918.73.
Filing withdrawals as income is the other tempting answer and is worse:
it books an asset disposal as earnings and feeds the same figure into net
with a flattering sign. Same reason the Up item sales in Known Gaps do
not belong on the income line.
What this cannot resolve: part of a withdrawal genuinely is income — the
capital gain. The bank descriptor is one gross figure with no cost base,
so it cannot be decomposed from statement data. Netting tracks cash
committed against cash returned and leaves the gain for holdings data to
surface; it does not assert the gain is zero.
The budget page gates the Invested card on `!== 0` rather than `> 0` — a
net-disinvesting month is real data, not an empty one — and renders
negative months in amber so the sign is not hidden by matching digits.
Transfers move money between your own accounts; at 433 of 3,996 rows
(~11%) they crowd out the rows that represent actual spending.
getTransactions gains `exclude_categories`, opt-in per caller and
deliberately not defaulted in queries.ts: the rules preview and the bulk
rule-apply path both read candidate rows through getTransactions, and a
default exclusion there would silently shrink what a rule can see and
reach — invisibly, since a rule that matches nothing looks the same as a
rule with nothing to do.
Two behaviours the filter needs, both tested:
- An explicit category pick beats the exclusion. Selecting "Transfers"
while the default is on subtracts it from the hidden list instead of
returning zero rows and reading as "you have no transfers".
- COALESCE the effective category to '' before `<> ALL`. NULL <> ALL(...)
is NULL, not true, so an uncategorised row would disappear from a
filter that never named its category — the trap EXCLUDE_NON_SPEND
already documents.
The default is off when the view is scoped to a statement: that is a
reconciliation view, the row count has to match the statement, and a
credit-card payment is the row you went there to check.
Status for this app now lives on the Vikunja board (saved filter
finance-app), which replaced the smarthome repo's ACTIONS.md on 2026-07-30.
Notes the thing a single-label filter hides: a ticket can carry several
system labels — the receipt→pantry work is finance-app, pantry-app and
email-ingestion at once — so the finance filter is a view, not the boundary
of what will touch this codebase.
Also records the one dated item here: postgres-personal runs PostgreSQL 14,
EOL 2026-11-12, and it holds statements, transactions, orders and
expense_metadata.
The order-details panel reads qty/description/amount from
expense_metadata.line_items. A grocery shop is a receipt like any other, so it
stores the same keys rather than name/quantity/line_total — otherwise the rows
arrive complete and display blank, which is exactly the failure ING-8 names.
unit and category are the two fields a grocery line has and a delivery line does
not; nothing renders them yet and the category composition will.
Adds /api/receipts/ingest as a sibling to the order lane, sharing its shape but
making one decision differently: nothing is parked. An order can wait for its
statement because it is already visible as an email; a gift-card grocery shop is
visible nowhere at all, so a scan that produces no transaction produces nothing
anyone can see. Every payment becomes a manual row immediately and the existing
pending-reconciliation queue resolves the ones with a card leg coming.
One transaction per tender leg. A $114.57 shop settled $40.75 gift card +
$73.82 Mastercard has a statement line for $73.82 only. A single row marked
credits is excluded from the queue while that line double-counts; marked card it
is searched for at 1% of $114.57 and never matches. Either way the shop books
$188.39. Per-leg rows make each amount the settled amount, so the matcher works
untouched.
Reconciliation now carries expense_metadata across. It already moved overrides,
tags and splits from the manual row to the statement row and left metadata
behind, which did not matter while metadata only came from an email that made
its own transaction. It matters now that it carries a shop's line items:
unmoved, the contents vanish at exactly the moment the statement line appears,
and COLES 0556 MANOR LAKES stays as unreadable as before anything was scanned.
transaction_id is UNIQUE, so a statement row that already has metadata keeps it
and the pantry row is flagged rather than raising a constraint violation.
Also regenerates the Prisma model. card_last4, currency, flags, reconciled_at,
matched_transaction_id, platform and route have been in the database since
migrations 0019/0020 and were absent from schema.prisma — regenerating the
client from it would have dropped columns the order lane writes on every ingest.
23 integration tests against the real schema, built from the three receipts that
drove the design. Existing suites unchanged: 104 unit, 144 integration.
The redesign is live (2026-07-28); only the loan model remains a proposal.
The UI review's 'settled is dead data' guidance described the pre-rebuild
state and is superseded — settled now gates ACTIVE_OBLIGATION.
A 50/50 arrangement was stored as a single row saying "Sonu 50%". The
arithmetic was never wrong — `myShare` resolves the payer's share as
`100 - SUM(everyone else)`, so balances and per-user spend were correct
throughout. It was still a bug, because a ledger is read as well as
computed: on screen that row is a 50% share against a blank, which looks
like half the money is unallocated and is indistinguishable from a split
somebody abandoned half-finished.
It also leaked. `getSharedTransactions` filters by participant with an
EXISTS on an explicit split row, so filtering the Shared view by the
payer silently dropped every transaction where their share was only ever
implied.
Four write paths could produce it, three of them unguarded:
- the Slack nudge's share button, which inserted one row
- `POST /api/transactions`, where the add form shows an amber total
under 100 but saves anyway — this is how Lawn Mowing and Hedge
Pruning were stored
- `applyRuleActions`, where ten of the fourteen live split rules name
only the other person
`completeSplit` is now the single place that writes the remainder, and
every one of those paths ends in it. The remainder goes to the
transaction's *owner*, never to "me": the owner's row on their own
transaction is excluded from both halves of the balance query, so it
cannot create, enlarge or discharge a debt, whereas a row for me on
someone else's transaction is a real obligation. That distinction is
what makes this safe to apply to existing data.
Also fixes the order panel's "Shared 50/50" toggle, which was inert in
both directions: it posted a lone 50% row to share (rejected — must
total 100%) and an empty array to un-share (rejected — array required),
because no way to clear a split existed. DELETE on the splits route is
that way.
Backfill: 7 rows, verified against a row-level dump diff — 2549 -> 2556
rows, none removed, none modified — and participant balances byte
identical before and after (Molina 19556.07, Sonu 20913.35). Every split
in the database now totals 100%.
Not done: a database-level constraint. Enforcing the sum needs a
deferred constraint trigger, and the rule path commits its DELETE and
INSERT as separate statements, so the trigger would reject the
intermediate state. Making it work means wrapping every write path in a
transaction, which is a larger change than the defect warrants.
`•` was missing from the entity table, and that was not cosmetic.
DoorDash separates an item's name from its options with a bullet and
parseDoorDashLineItems splits on the literal "•" — so left encoded, the
split never happened and the line collapsed into the description:
"Bucket and Side Pack (Meal Deals) • Hot Bucket • Chips" with
options []. The entity showed on screen and the structure behind it was
gone.
Numeric entities are now decoded generically rather than one at a time,
which is how $ came to be listed individually while its neighbours
were not, and & resolves last so a literal "&bull;" stays as
written instead of turning into a bullet.
Repaired the 47 stored rows by re-parsing the captured email behind each
one rather than string-replacing the entity, since a replacement would
have fixed the display and left options [] underneath. Rehearsed first:
all 47 re-parsed, all 47 gained options, 0 line items lost, 0 amounts
changed. Old values kept in dump/rollback-line-items-20260728-223737.json.
Pre-existing — 22 rows predate today — but the pre-cutover backfill more
than doubled the affected rows, which is what surfaced it. Verified in
the Order details panel, not in SQL.
I1 refused any credits-funded order dated before 2026-01-09, storing
nothing at all — no transaction and no metadata, so the receipt was
discarded rather than kept as history.
Its reason was splits, not spend: before the cutover shared expenses
lived in SplitMyExpenses, and re-importing them would double-charge
against carryover transaction 2348. That reason expired with 788219b,
where ACTIVE_OBLIGATION became `settled = false AND transaction_date >=
'2026-01-09'`. A pre-cutover split can no longer assert a debt, so a
pre-cutover order cannot move a balance however it is recorded — and
ingestion writes no splits at any date, which now has a test of its own.
What the guard was still doing was hiding ordinary history: 275 orders,
$9,799.96 of meals and rides across 2020-2025, invisible only because the
money came from a gift-card balance instead of a card.
The funding side stays as it is, deliberately. Some of those orders were
paid from ShopBack gift cards that are themselves booked as expenses, so
that portion is counted twice. The exposure is bounded at $3,411.16 over
14 loads and is probably smaller: the descriptors name no brand — "ShopBack
Gift Cards SQ" is a batch code, and the card could be Amazon, Airbnb or
Shell as easily as DoorDash — and six are categorised `gifts`, which may
be real presents rather than self-funding. Reclassifying them on a guess
would corrupt correct data to fix a double-count that cannot be shown.
Only the ShopBack purchase emails can settle it, joined on total paid.
Uber Cash IS credits, and the payment line had become unreadable in the
newer layout: "Payments Uber Cash 10/17/25 8:50 PM A$54.87" carries a
timestamp between the label and the amount, and writes the currency as a
prefix. Both defeated the pattern, so credits_amount stayed null and the
order was filed as card-settled — sent looking for a card leg that does
not exist, found nothing, and left as an orphan with no transaction and
no card tail to match on. 118 captured messages sit in that state, and
every one of them is pre-cutover, so reading them correctly means I1
skips them rather than storing enrichment that points at nothing.
The card leg had the same blind spot, hidden behind the first: the gap
between the mask and the amount was [^\d]{0,40}, which a timestamp
breaks. While BOTH legs were unreadable a mixed payment still looked
consistent — the order read as card-settled for the full total. Fixing
only credits reads half an order, which validateOrderTotals correctly
refuses. Found exactly that way: four messages that validated before
began failing "payments sum to 34.92 but receipt states 43.60".
Legs are summed rather than taken first, because one order can be charged
in instalments and an instrument can carry no mask at all (PayPal). But a
leg that already equals the stated total IS the payment, not an
instalment: a Dubai trip prints an AED 17.67 authorisation and then the
AED 577.83 settled charge, and adding the hold overstates the trip. A
mixed credits+card order is unaffected — neither leg equals the total
there, which is why it needs summing.
A/B over all 776 captures: 468 parsed by both, zero change to any amount,
currency or existing card tail, 0 lost, 118 credits figures newly read.
Mutation-tested: dropping the exact-leg rule fails 2 tests, dropping the
Uber Cash read fails 7.
Two parse bugs that between them made 218 of 776 captured messages
unreadable. Neither was the "old template" they were filed as.
Uber writes the currency three ways and only two were handled. "Total
A$54.87" is what it sends for ordinary Australian orders — A$ misses
[A-Z]{3} by one character — so 98 of 275 Uber Eats mails and 78 of 275
trip mails failed with "no Total found" while the amount sat in plain
sight. Most were 2024-2025, i.e. current mail. NZ$, US$, S$, HK$, C$ and
a bare rupee/euro/pound symbol are handled the same way. A bare "$" is
still left unresolved on purpose: a dozen currencies use it, and the
body-wide scan that reads the receipt's own stated code should win.
A DoorDash order paid from credits states "Total Charged $0.00"
truthfully, above a real subtotal. Read literally that is a $0 order, and
validateOrderTotals rejected 88 of them as non-positive — discarding
exactly the credit-funded spend this pipeline exists to surface. The
order's value is its subtotal; recording zero would show the order and
hide what it cost. Guarded on the receipt actually saying credits, so an
empty mail still fails rather than inheriting a stray subtotal, and the
header cross-check stands down for these or it rejects the figure the
parser deliberately overrode.
Measured A/B over all 776 captures: 254 parsed by both parsers with zero
change to any amount or currency, 0 lost, 214 newly readable. Both fixes
mutation-tested — reverting the regex fails 4 tests, removing the credits
branch fails 3.
Fixtures are real captured receipts, per the 2026-07-26 rewrite: the
earlier synthetic suite passed while the parser could not read a real
email.
No history replay — I7 idempotency refuses re-reads and that needs an
explicit update mode. This fixes ingest from here on.
The jump from "OK" to "Never again" is too big and most
disappointments live in the gap (user, 2026-07-28) — so a merely poor
meal either flattered itself as OK or got blacklisted.
Only 'never' raises the warning on a future order. A blacklist that fires
for every mediocre delivery is one nobody reads, so 'bad' records the
disappointment without triggering the alarm. Both set order_again = false
— you would not choose either again — and that split between "would I
order it" and "warn me about it" is the point of the extra level.
Migration widens the CHECK; nothing is removed, so no existing row needs
mapping.
Every press wrote correctly and then left the card showing stale state,
so a working button looked dead — and a button that looks dead gets
pressed again, which toggled the split back. That is how two orders got
unshared while looking like nothing had happened.
The cause was a wrong assumption in the original design: Block Kit
interactivity ignores the HTTP response body. Replacing a message from
the response is legacy attachment-style behaviour. The update has to go
to payload.response_url, which needs no bot token — so it stays in the
app rather than becoming another n8n node.
Two things that were also invisible now speak up. The share guard used to
return silently when it refused a three-way or uneven split, which is
indistinguishable from a broken button; it now says which it was, as an
ephemeral only the presser sees. Same for an unmapped Slack user.
The response still echoes the rendered blocks. Slack ignores them, but it
lets a card be rendered server-side without pressing anything — which is
what stops the replay tooling from hand-writing a card with a guessed
share state, the mistake that cost a real split earlier today.
Two halves of the same requirement, one of which was quietly missing.
Sharing split the money but never reached her: she is not in #smarthome,
so the card whose caption said "both verdicts welcome" was one she could
not see. Now a share DMs her a card of her own. A DM rather than adding
her to the channel, so her surface stays "orders that concern me"
instead of the whole house's ops feed. She was already in SLACK_USER_MAP,
so her press files under participant 4.
Only on the press that turns sharing ON, and only when someone else did
the sharing. Re-notifying on every later rating press would turn one
shared meal into a stream of DMs, which is how a nudge gets muted.
Her card carries no share button: she is being told it was shared, not
asked to decide, and two people toggling one split from separate copies
of a card is a race with no upside.
The app still holds no Slack bot token — it returns a notify instruction
and n8n sends it, the same shape as the modal open. If SLACK_USER_MAP has
no id for her the DM is skipped silently: the split is correct and
complete either way, and failing the press over an unaddressable nudge
would be the worse trade.
Card reordered to rate -> details -> share. You judge the food, then
decide who pays for it; asking "was this shared?" first inverts the
order a person thinks in. The status caption moved under the share button
it describes rather than sitting orphaned mid-card.
Reverses a change made on request. The platform in the headline
fragmented the merchant: the same restaurant read as two, depending on
who carried the bag, and that is not a distinction anyone rating the food
cares about. It also already has a home — the expandable Order details
panel renders expense_metadata.platform next to its heading, which is
where the user asked for it.
The fragmentation was worse than cosmetic. merchantVerdict joined on an
exact merchant_normalized, and the platforms capitalise differently
("TEG Kebabs & Biryani" on Uber Eats, "TEG KEBABS & BIRYANI" on
DoorDash), so one restaurant kept two separate histories and a "never
again" recorded through one app never warned in the other — silently
defeating the point of the memory. Now case-folded; verified on real
data, where the DoorDash order sees 1 prior verdict against 0 before.
81 existing descriptions backfilled in one transaction, dry-run first and
dumped beforehand. The regex is anchored to the end so suburb parens
survive: "Order - Coles (Wyndham Vale) (Uber Eats)" becomes
"Order - Coles (Wyndham Vale)", not "Order - Coles".
Splits on orders are made by hand, so a third participant or an uneven
share is a deliberate decision — and "Make it just me" deleted every
split row regardless. A one-tap button silently destroying an
arrangement made with more care than the tap that undid it is the same
failure shape as the rewrite that dropped `settled`.
Now it refuses when a participant other than the two consumers is
present, or when the share is not 50. Verified against the running stack:
a three-way split and a 70/30 both survive a press; a plain 50/50 still
toggles off and back on.
Also: the nudge reads share state instead of assuming false. Today a
freshly ingested order has no splits — the 140 that do were split by hand
after the backfill, not by a rule — but the label drives a destructive
button, so a wrong assumption there costs data rather than a cosmetic
error. One query is cheaper.
The card can rate an order but cannot ask which dish or why: a message
cannot collect free text, and an actions block caps at 25 elements while
item counts vary per receipt. A modal is the only Slack-native answer,
and it stays inside Slack — no browser, no app, which is the whole reason
it exists rather than a link.
The overall rating deliberately stays on the card. That is the thing done
every time and it should cost one tap; this is for when something was
notably good or bad.
finance-app holds no Slack bot token by design, so it returns the view
and n8n — which already has the credential — calls views.open. One copy
of the token, no new secret, no compose change.
Item text travels in private_metadata because a submission returns block
ids and values, never labels, so there is otherwise no way back to which
dish a radio button referred to. Capped at 20 rows: a grocery order runs
long and nobody scrolls a modal to rate a tin of tomatoes.
The modal does NOT write the rating. A form that silently reset a
decision the user did not revisit is the same class of bug as the split
rewrite that dropped `settled`.
Slack's mobile client gives every button in an actions block its own
full-width row, so the four ratings rendered as four stacked bars and the
nudge filled the screen (user, with a screenshot). A select is one row
and still one decision.
The interactive route now resolves both shapes — a button carries
`value`, a select carries it on `selected_option`. Reading only the
former would have left sharing working while rating silently did
nothing.
Being sent to a web app to answer "was this shared?" is enough friction
that the question stops getting answered — which is the exact failure the
nudge exists to prevent. So the buttons now act in place: pressing
"Shared 50/50" writes the transaction_splits row and edits the message,
and the app is never opened.
Slack does NOT reach this route directly. It posts to an n8n webhook that
forwards the raw body and signature headers here (user's suggestion).
That is the better shape: n8n already terminates public webhooks, so the
app keeps its blanket OAuth chain and gains no internet-facing
unauthenticated route, and no Traefik change is needed. n8n cannot verify
the signature itself — its Code sandbox has no `require`, so no `crypto`.
Two independent gates, both failing closed: the shared x-ingest-token
(came from n8n) and Slack's v0 signature over the raw body (came from
Slack, not replayed within 5 minutes). An unset signing secret rejects
everything rather than waving it through, because the alternative turns a
misconfigured deploy into an open write endpoint.
An unmapped Slack user is refused rather than defaulted to the owner. In
a two-person household a wrong attribution is not a rounding error, it is
the other person's opinion recorded under your name.
Block Kit is built in the app (ingest returns slack_blocks) rather than
in n8n expressions: a template string is untestable, and this shape has
to stay in step with what the interactive endpoint re-renders after each
press. Null when there is no transaction yet — a card-settled order is
parked until its statement arrives, so there is nothing to split or rate.
Also: /transactions now honours ?q=, so the link lands on the row instead
of the top of an unfiltered ledger.
The Slack order nudge links here. Without it the link drops you at the
top of an unfiltered ledger and the merchant has to be found by hand,
which is how a nudge stops being opened.
The ledger already knew we had ordered from a place; it did not know the
food was bad. Orders got repeated from places we disliked because nobody
remembered by the time the next one went in. That is what the receipt
ingestion was for (ING-9) and the last piece was missing: order_reviews
existed as a table with no API, no UI and no writes.
Four levels, not three. "Loved" and "liked" are both "would order again"
but only one is worth a detour, and "ok" is not a recommendation.
A verdict belongs to a person, not to an order. A shared meal produces
two opinions and they routinely disagree — that disagreement is the
useful part, and the old UNIQUE on transaction_id alone could not hold
it. Now UNIQUE (transaction_id, participant_id), and the default is the
signed-in user rather than the owner: Sonu authenticates through the same
Traefik OAuth as participant 4, so an owner default would have filed her
verdict under his name.
Per-item opinions key on the item DESCRIPTION, not its index. An index is
meaningless across orders; "the Pad Thai here is good" is the signal that
has to survive into the next order from the same merchant. Only the two
poles are offered — a per-item "ok" answers neither of the questions you
ask at order time.
Sharing is recorded as a real 50/50 split, not a decorative flag. The
split already IS the record that an order was shared, and two records of
one fact drift apart.
An ABSENT item_verdicts means "leave them alone"; an empty array clears
them. Without that distinction a note-only save silently wipes every
per-item opinion — the same shape as the bug that reset `settled` on
split rewrites, and just as invisible on screen. Mutation-tested: making
keepItems a no-op fails exactly one test.
mockDbWithPool gained queryRow. Omitting an export from the mock makes it
undefined at the call site, which fails as "not a function" and reads
like a code bug rather than a test-harness gap.
Every warning in it was inverted by this week's work:
- "settled is dead data, false on every row" -- there are now 1,266 settled
splits across 657 pre-2026 transactions.
- "do not fix getParticipantBalances to exclude settled splits" -- it now
excludes them, via ACTIVE_OBLIGATION, and must.
- "settlement cannot be attributed per trip" -- migration 0022 added
split_payments.trip_id and it is attributed.
- "splits exist from 2026-01-09 only" -- pre-2026 transactions are now split
deliberately, to stop them inflating spend.
Replaced with what is actually true, including the rule that matters most: the
cutover DATE is the primary balance gate and the settled flag only refines it,
so pre-2026 expenses can be split freely.
Nothing dated before 2026-01-09 can be owed, because carryover transaction 2348
already carries the entire pre-cutover balance as a single figure. ACTIVE_OBLIGATION
now says so directly.
This inverts which mechanism is load-bearing, and that is the point. Until now
the only thing keeping $37,233.28 of paid debt out of the balances was
transaction_splits.settled -- a boolean that any delete-and-recreate write path
resets to false, as the split modal did until commit 6add958. Losing the flag on
a pre-cutover row now costs nothing: the date still excludes it. The flag matters
only on or after the cutover, marking the few settled outside this app.
It also makes splitting history safe to do freely. A split on a 2024 grocery
shop can now describe how the expense was shared -- which is what stops it
inflating spend -- without asserting a debt that was settled years ago. That was
the whole reason not to split pre-2026 expenses, and it no longer applies.
The bound is inclusive because transaction 2348 is itself dated 2026-01-09; an
exclusive one would drop the carryover and with it the entire pre-cutover
balance.
Corrects one live figure: a Woolworths on 2026-01-06 was split 50/50 three days
before the cutover, double-counting $7.55 against the carryover. Sonu
$5,428.08 -> $5,420.53.
The test fixture default moved to 2026-06-15 -- it was 2024-06-15, which is now
pre-cutover and made every balance fixture read zero. That the suite caught this
is the guard working.
The route replaces every split for a transaction rather than editing in place,
so the recreated rows took the column default settled=false. Opening the split
modal on a historical transaction and saving it therefore converted a
discharged obligation into a live one, with nothing on screen saying so.
That is not theoretical. 657 pre-2026 transactions now carry settled splits
imported from SplitMyExpenses -- $37,233.28 of balance that the carryover
(transaction 2348) already accounts for. Editing one would double-count its
share against a debt that was paid years ago.
Now carries settled and settled_at across the rewrite, per participant, the
same way the rules revert route already does. Changing someone's percentage
does not re-open the obligation: it was settled outside this app and stays
settled. A participant who was not on the transaction before is a genuinely new
obligation and correctly starts unsettled.
rule-actions.ts was already safe here -- it upserts ON CONFLICT DO UPDATE SET
share_percent, so it never touches the flag.
scripts/split_csv_match.py needs psycopg2, so scripts/ now has a venv beside it.
This repo deploys from its working tree, so an untracked .venv would be swept
into the Docker build context.
Sonu's two "transfer" payments are split by scope, Europe first and the
remainder to household, chronologically so each settles what was outstanding
when it was made. Both Europe tabs now read $0.00.
No schema change was needed and that is the point worth writing down:
split_payments has no unique constraint on linked_transaction_id, so one bank
transfer carries one row per scope and the rows re-add to the transfer --
verified, 4111 sums to $3,779.33 and 4121 to $4,794.06.
Her overall balance is unchanged at $5,428.08. Allocation moves money between
tabs, never between people; that invariance is the check to repeat on any
future re-allocation.
Trip 3, 2026-04-12 to 2026-04-28, 124 rows, $9,914.24. It was marked only by
tag 5 and so was invisible to every trip figure.
It reads at first like a sharing scope overlapping the group trip, because the
tag's earliest row is 17 March. It is not: that is a single advance booking
(Ticketmaster Nanterre), 8 more rows fall on the 12 April handover day and were
already held out of Europe 2026, and the remaining 115 run 13-28 April. A clean
sequential leg.
Also records that grouped payments need no schema change -- split_payments has
no unique constraint on linked_transaction_id, so one transfer can carry a row
per scope -- and that the 23 Apr Qantas booking is the flight to Bangkok
starting a solo leg, which is why it stays out of this trip.
The 2026-07-26 document was a proposal marked "nothing built". Everything it
described as broken is now fixed, and the fix is not the one it proposed, so
leaving it in place would misdescribe the system to whoever reads it next.
Records what the code now does: settled as the single balance gate, settled and
trip_id as orthogonal axes, settling up by recording a payment rather than
flipping a flag, and the reasons duplicates are superseded rather than deleted.
Keeps the loan design intact and clearly marked as still a proposal -- it was
never built and nothing in this work touched it.
Also records that the proposal's own recommendation not to restate history from
the CSVs was overturned, and why it was wrong: it measured the value in
balances, where it is nil, and missed it in spend, where it is $35,259.
Adds --write to the matcher. Wrote 1,242 split rows across 657 transactions.
Imported settled, and that is the whole design. These obligations were
discharged years ago on a platform we no longer run, and their residual is
already carried by transaction 2348. Writing them unsettled would re-open
roughly $40k of debts that were paid. ACTIVE_OBLIGATION keeps settled splits
out of every owed figure while myShare/mySplitOf still count them, which is
exactly the asymmetry this needs: the import exists to correct historical
SPEND, not to move a balance.
Effect: $35,259 leaves my historical spend -- $13,088 in 2024, $22,117 in
2025 -- because a $200 grocery shop that was always half hers no longer reads
as $200 of mine. Balances are byte-identical before and after (Molina
-1226.72/145, Sonu 5428.08/419), which is the assertion that matters.
Shares are written as the CSV computed them, so a 50/50 row can land as
50.01/49.99. That is faithful rather than tidy; no transaction exceeds 100%.
Rehearsed on the 37-row Rome file first (24 rows) and verified before the full
run -- both the balances and one split read back through the API.
Matches the five CSV exports against transactions already in the ledger and
reports what it would do. Writes nothing -- importing is a separate step, and
rehearsing it first is what catches the defects that tests do not.
What it found, and why the number is what it is: 676 of 1,536 shareable rows
match (44%). The ceiling is ledger coverage, not matcher quality. The CSVs
describe 678 shared expenses in 2024 alone; the ledger holds 591 rows for the
whole of that year, 3 to 72 a month, which is far less than a household
actually spends. Most 2024 CSV rows have no transaction to attach a split to
and never will. South Korea April 2024 matches 4 of 158 for that reason.
Three decisions are encoded deliberately:
- Date format is decided per FILE, not per row. The household export writes
D/M/YYYY and the four trip exports write ISO, and 474 rows parse validly
under both readings -- per-row guessing silently swaps January and February
for some rows and not others.
- A person's column is their net balance impact, not their share. The payer
is whoever is positive; the other's share is |their negative| / cost. So a
+cost/-cost row means the other party owes 100%, not that the expense was
unshared -- the reading that would fake an arrangement change.
- Matching is one-to-one, best pair first. The NZ trip has two identical
$10.16 Uber rows against three ledger rows and four PayMyPark rows in the
same shape; without this a ledger row is claimed repeatedly and the second
CSV row looks matched while being unrepresented.
Statements 107, 142 and 143 bill overlapping periods on one ANZ account, so 31
transactions -- $42,040.68 -- are in the ledger twice.
They are marked superseded, not deleted. Every child of transactions is ON
DELETE CASCADE (splits, tags, overrides, expense_metadata, order_reviews), so
deleting "the duplicate" destroys whatever curation sits on it, and which
member of a pair holds that curation is an accident of import order: here 1
pair carries splits and 6 carry overrides, all on the surviving side, but
nothing guarantees that. Superseding keeps the row, keeps its children, and
makes a mistake one UPDATE to undo rather than a restore from backup.
reconciled_with_id could not be reused. Its predicate is scoped to
statement_id IS NULL on purpose -- a statement line pointing at something else
is the survivor, not the duplicate -- and here both rows are statement lines.
The exclusion goes into EXCLUDE_RECONCILED_SOURCE rather than into a new
fragment, so every query already asking "count each purchase once" gets it
without being edited. The trip cost queries did not use that fragment at all
and now do; verified a no-op on current data (0 trip-tagged rows are either
reconciled sources or duplicates), but they were one import away from
double-counting.
Most of the $42k is transfers and investments, which spend already excludes.
The damage was elsewhere: duplicated rows in the list, and rules re-splitting a
duplicate -- txn 3807 is one of these 31 and was a candidate for splitting
earlier today.
Balances are unchanged: no duplicate carried a split.
An account cannot be billed twice for the same day, so an overlap means those
transactions are in the ledger twice. ANZ statements 107 and 143 overlap by 118
days and put roughly $42,000 of duplicate rows in; nothing anywhere said so.
Two details decide whether this catches the real case:
- Account numbers compare with non-digits stripped. The duplicate got in
because the existing key compared raw text and ANZ wrote the same account
as 408556264 on one statement and 4085-56264 on the other.
- The range is half-open. These statements are issued back-to-back with one
period ending the day the next starts, so inclusive bounds flagged 5 pairs
of which 3 were consecutive and fine. Half-open leaves exactly the 2 real
ones.
NULL bounds are excluded rather than handed to daterange, where NULL means
unbounded and an undated statement would overlap all of history.
Detection only. It does not refuse the import or touch the duplicate rows --
cleaning those is separate, and must supersede rather than delete because every
child of transactions is ON DELETE CASCADE and the curation sits on the
duplicate side.
Both subtleties have a test, and both fail if you undo them.
A participant who has overpaid a trip showed as "$-816.16" under a column
headed "Outstanding on this trip". A negative outstanding reads as a bug
rather than as "they are ahead", so the sign is now spelled out: magnitude
plus one of all square / owes you / ahead — you owe them, coloured the same
way Shared colours the same three states.
Also corrects the footer, which had gone stale and was now simply false. It
said settlement could not be computed per trip because payments carried no
trip attribution. Migration 0022 added split_payments.trip_id and the figures
above it have been net of trip-scoped payments since. What a reader needs to
know is the opposite of what it said: household-tab payments are the ones NOT
counted here.
Total Spend is every payer's trip-tagged spending; the split figures directly
below it are scoped to the owner. Two lenses on one screen read as one unless
the card says which it is. The number is unchanged and deliberate -- a trip
cost what the group put into it -- so this is a label, not a fix.
Every trip figure filtered on transaction_type IN ('debit','fee','interest'),
which drops refunds and credits outright. A partly-refunded booking therefore
read at its full price and the refund subtracted nothing, anywhere: the
headline total_spend, the category breakdown, the daily chart, top merchants
and the tag breakdown were all gross.
This is the same defect the general analytics fixed once already, which is why
NET_SPEND_ROWS and SPEND_SIGNED exist -- a refunded Expedia purchase read as
$2,888.92 of spend until they did. Trip analytics never adopted them. Doing so
now costs one predicate and one expression per query.
getTrips/getTripById needed the trips alias moved to `tr`: the fragments assume
`t` is `transactions`, and hand-inlining a copy rather than renaming is exactly
how the reconciled-row exclusion drifted out of the analytics routes before.
On Europe 2026 this is $821.12 -- a LuxuryEscapes booking with two part-credits
against it, and a FreeNow hold adjustment. Fully cancelled bookings are a
different case and are handled by untagging both legs from the trip by hand,
because a trip never incurred a cost it cancelled.
No balance moves: the owed query already excludes credits and a refund carries
no split. There is a test asserting exactly that, and it passes with or without
this change -- it is a guard, not a proof. The three that do prove it fail
without it.
The per-trip owed number shipped in 689fadc counted every split on every
trip transaction regardless of who paid, so it silently mixed debts owed to
different people under one label.
On Europe 2026 that meant Molina "owed" $21,572.12, of which $1,605.49 was
her share of rows Sonu paid for — a real debt, but between the other two
participants, and one they had already settled directly (split_payments id
5, Molina -> Sonu, exactly $1,605.49). A participant's own share of a row
they themselves paid for was in there too, which is nobody's debt at all.
Both sides needed scoping, not just one: the owed side to rows this owner
paid for, and the paid side to payments made to this owner. Scoping only
the first would have let a Molina -> Sonu payment reduce what Molina owes
the owner.
The corrected figures reproduce a number derived independently, months of
data apart: Molina now reads -$816.16 on Europe, matching her known
overpayment to the cent ($19,966.63 of splits against $20,782.79 paid).
Sonu goes from $8,793.10 to $1,084.61, and the owner correctly disappears
from a list of people who owe the owner.
Found by checking a household total against what the app had been showing
all along — the query was gross, gave a number about twice the real one,
and I had quoted it as "owed". Worth stating plainly: the defect was not in
the number the app displayed, it was in the number I computed to explain it.
The pipeline has been red on every run since at least ae0c34f. npm ci
installs dependencies but the Prisma client is generated into
src/generated/prisma, which is gitignored — so a fresh CI checkout has no
client and anything importing src/lib/db.ts fails with 'Cannot find package
@/generated/prisma/client' before a single assertion runs.
prisma generate reads only the schema, so it needs no database and no
secrets.
Worth noting why this went unnoticed for a dozen commits: a pipeline that
is always red carries no signal, so it stopped being read.
a4ab543 landed six hours ago and this branch rewrote one of the queries it
had just fixed, quietly dropping both of its guarantees.
That commit made EXCLUDE_RECONCILED_SOURCE "one fragment both sides import"
because an inlined copy is how the reconciled-row exclusion drifted out of
the analytics routes and double-counted 48 rows / $4,474.79. The trip owed
query here had hand-inlined its own copy — the fragment assumes the alias
`t` and this query used `tx`, so the path of least resistance was to
re-create exactly the divergence that was being removed. Aliased to `t` so
the fragments apply directly.
The same commit made balances count rows whose AUD value is unknown rather
than netting a foreign figure against AUD ones. The trip figure had no
equivalent — on the query where it matters most, because a trip is where
foreign rows actually live. A Europe total silently mixing EUR into AUD is
the whole failure that fix was written to prevent.
The column header still read "Share of this trip" while the number is now
net of payments, which is the same class of drift a4ab543 set out to fix.
It reads "Outstanding on this trip", carries the approx/unconverted caveat
the Shared cards use, and greys a settled zero.
A payment has only ever recorded from, to, amount and date. That is why
the per-trip owed figure did not exist — getTripAnalytics said so where
the number should have been: "split_payments carries no trip attribution,
so a payment cannot be assigned to a trip. Settlement is a property of the
whole relationship." Every trip therefore read 100% unsettled, including
trips paid in full.
It is also why the Shared page silently drops payments under a tag filter.
With one global pool there was nothing honest to subtract, so it showed
gross splits under the same label. A tag is a view; a scope is a ledger.
The scope is a trip, not a new settlement_contexts table. trips already
has owner_id, dates and archived, and transaction_overrides.trip_id
already decides membership. A second grouping beside it would be two
unsynchronised scopes over the same rows, with no invariant saying which
governs. NULL means the ongoing household tab, which never closes.
settled answers a different question and the two must not be collapsed:
trip_id is which tab, settled is whether the obligation is still live.
Critically, a live obligation is NOT settled by flipping the flag — it is
settled by recording the payment, and the balance nets to zero on its own.
Doing both would subtract the settlement twice. So settled is written only
by the historical import, for repayments made on a platform we no longer
run, and there is deliberately no "mark settled" action.
Both owed figures now exclude settled splits and the trip figure nets its
own payments. Spend analytics (myShare/mySplitOf) deliberately still count
settled rows: my half of a 2025 grocery shop is my spend whether or not the
other half was ever repaid, and filtering them would re-inflate exactly the
figures importing settled history exists to correct.
Also drops /api/participants/[id]/balance. It had no consumers, no owner
scoping, no debit/credit signs and no EXCLUDE_RECONCILED_SOURCE — a fourth
balance implementation that disagreed with the others and would have
imported three bugs if anything had aligned to it.
getTripAnalytics had no test at all. It has five now, including the one
that matters: a household payment must not make a trip look paid. Verified
by mutation — neutering the settled filter fails three, and dropping the
trip filter on payments fails that one.
Six metric-integrity defects from the UI/IA review, plus two found while
verifying the review's own claims against the code.
The reconciled-row exclusion existed only in queries.ts. Every analytics
route counted the superseded manual rows as spend — 48 rows, $4,474.79 of
double count, invisible precisely because the transaction list looked
right. It is now one fragment both sides import.
The spend-pace chart computed its own totals in the browser: gross
amounts, debits only, no personal share, no refunds, fees, interest or
itemised loan repayments. On live data it ended July at $4,747.31 under a
headline reading $3,597.10 — and its own baseline line was drawn from the
split-adjusted monthly totals, so the two series in one chart disagreed
with each other. Both now come from /api/analytics/daily, built from the
same fragments as the headline.
Fees aggregated every statement ever imported with no date filter, under
a heading with no period, so a lifetime figure read as a current one and
grew forever. Now bounded, labelled, and selectable.
Comparisons no longer measure a month in progress against complete ones:
the in-progress month is out of every baseline, and a selected current
month is compared through the same day.
Two the review did not catch:
- Every analytics window was a day early. toISOString() on a
local-midnight Date converts backwards through UTC. Surfaced only once
fees started reporting the range it had used.
- /monthly rounded per category, /daily per category-day, so the pace
chart ended a few cents off the headline above it.
Shared currency needed amending rather than applying. Reading s.currency
would have labelled every order row AUD, since an order receipt has no
statement and carries its own currency — the opposite convention from a
foreign charge on an AUD statement, where amount IS AUD. NATIVE_CURRENCY's
COALESCE order keeps the two apart. Balances also now count rows whose AUD
value is genuinely unknown instead of netting a foreign figure against AUD
ones. Latent today: no foreign transaction is currently split.
Tag-filtered balance cards no longer claim "owes you". With a filter on,
payments are deliberately not subtracted, so the figure is a split total
and settling against it would record a payment for a debt that never was.
Split-coverage warnings deliberately omitted (user decision).
Enrichment is the point of the ingestion pipeline (DECISIONS ING-8) — a bank
statement gives a date, an amount and a mangled descriptor, and everything that
makes a transaction understandable arrives by email. It was all reachable only
by opening the edit modal, which is a strange place to look for "what was in
this order".
Rows with a receipt behind them get a disclosure arrow in the description cell;
clicking expands an inline panel with the line items, the pick-up and drop-off
stops, the card tail and the provider's reference. Several rows can be open at
once — the point is comparing orders without losing your place.
The arrow appears only where `order_platform` is set. Putting one on every
transaction would promise detail that mostly does not exist.
OrderDetails moves out of edit-transaction-modal.tsx into its own component so
both surfaces render the same thing; `bare` drops the modal's top border when
it sits in a table row.
Five rows all reading "Order - Uber Trip" are indistinguishable — the list gives
you a date and an amount and nothing to tell one ride from another (user,
2026-07-27). Where the trip went is exactly what separates them, and it was
already stored on expense_metadata.route since this morning; nothing in the list
read it.
getTransactions now joins the receipt (both directions — transaction_id OR
matched_transaction_id, since a card-settled order points at the statement line
instead) and the description cell renders "Terminal 2, Melbourne Airport (MEL)
→ 19 Lady Penrhyn Dr" in the same italic sub-line notes use.
Two deliberate limits:
- **A note the user wrote always wins.** This only fills an empty sub-line; it
never occupies the notes field, which is theirs.
- **Deliveries are excluded.** Their merchant already identifies them, so the
restaurant's street address would be clutter on every food order. Gated on
platform = 'uber'.
The summary keeps the first two comma-segments of each address — a truncation,
not a guess about geography. Uber puts the venue or street first, which is the
identifying part; the full stops with their times stay in the title attribute.
Trip rows read "Order - Uber Trip (Uber)". The suffix exists so you can tell
where to go and look; when the merchant is literally "Uber Trip" it says
nothing. What identifies a trip is its two addresses, and those are in the
Order details panel. Existing rows updated in prod.
Local rides are paid with credits (only overseas ones go on a card), so trips
belong to this slice and were simply never fetched — the Graph query searched
"order with Uber", the Eats subject. Captured 15 real messages from the mailbox
via a dry-run before touching anything, which found two defects that no amount
of reasoning about the template would have:
**Uber sends two mails per trip.** A "charge summary" when the ride ends, then
the real receipt when payment settles — same subject, same total. The summary
carries no tripReference, so order_reference fell back to `msg:<message-id>`
and I7 could not dedupe it against the receipt that follows. Every trip would
have been recorded twice. It says so itself ("This is not a payment receipt ...
You will receive a trip receipt when the payment is processed"), so it is now a
NotAReceiptError — 200 and silent, like every other expected non-receipt.
**Trip receipts label neither end of the journey.** Delivery receipts write
"1:20 pm - Pick-up"; trips print the time alone. The split regex put the time
into `label` and left `time` null. Time is now read properly, and a two-stop
trip is labelled Pick-up/Drop-off positionally — only where the receipt was
silent, so a template that does label its stops keeps its own wording.
Verified against all 15 captured messages: 7 trips recorded, 5 charge summaries
and 3 promotions skipped, 0 failures, no duplicate references. Two of the seven
are AUD credits-funded ($84.78 + $47.97) and would become transactions; the
five NZD ones are card-settled and correctly create provenance only (I5).
Fixtures ut-00 (local credits trip), ut-01 (overseas card trip) and ut-summary
(the charge summary) are captured mail, not written by hand.
These tests insert a Westpac statement and a `DD *DOORDASH ...` charge, and
only removed them at the end of the test — so they survived into the next run,
where `reconcileCardLeg` could match one at ingest time and resolve an order
that was meant to park `awaiting_card_statement`.
That is a real ordering bug in the fixtures regardless. It is my best
explanation for the intermittent failure in "parks an unresolvable split",
but I could not reproduce it: seeding the exact leftover row and running the
old code passed anyway. So this is hygiene with a plausible mechanism, not a
confirmed fix — if that test fails again, this was not the cause.
`expense_metadata` has held the itemised receipt since ingestion started and
nothing in the UI ever read it. A transaction that came from a DoorDash or Uber
Eats receipt showed a merchant and an amount, with the item list and the
delivery addresses sitting unread in the row behind it (user, 2026-07-27).
Adds GET /api/transactions/[id]/order and an "Order details" section in the
edit modal: line items with their options, pick-up/delivery stops with times
and addresses, the card tail when one was involved, and the provider's own
order reference.
Two details that matter:
- The lookup resolves from **both** sides — `transaction_id` OR
`matched_transaction_id`. A card-settled order creates no transaction of its
own (I5); the receipt points at the statement line instead. Matching only on
transaction_id would have left the panel blank on exactly the card-paid
orders, which are the ones whose detail is hardest to find elsewhere.
- An empty item list says so in words rather than rendering nothing. Uber
itemises groceries but not restaurant orders, and orders ingested before the
Uber item parser existed have none either — a blank section reads as a bug
when it is usually the receipt.
Read-only. This is what a provider sent; editing it would make provenance mean
nothing.
Four things the view could not tell you, all from reading the rows (user,
2026-07-27).
**Which platform.** The parser has always known — it has to, to read the
template — and then discarded it. "Order - Burger Corner" gives no way to know
whether to open DoorDash or Uber Eats for the detail, and restaurants exist on
both. Now stored on expense_metadata and named in the description:
"Order - Burger Corner (Uber Eats)". Migration 0021 recovers it for the 101
backfilled rows from the order_reference shape — DoorDash receipts carry no id
of their own so ingestion synthesises `msg:<message-id>`, Uber carries a real
trip UUID, which makes the discriminator exact.
**Bank said "Manual".** That label is derived, not stored, and "Manual" reads
as "hand-entered, still awaiting a card line to match". A gift-card order has
no card line coming, ever. It now reads "Gift Card", and — the part that
actually mattered — credits joins cash in needsCardMatch(), so these stop
sitting in the pending-reconciliation queue. All 81 were queued against a match
that could not exist.
**Uber line items were never parsed.** 67 of 101 orders had none. Uber itemises
groceries but not restaurant orders, so some of that is genuine; the rest was
simply unread. Its markup is better than DoorDash's — every cell carries a
data-testid with the item's uuid, so qty/title/amount bind by id rather than by
column position. Sold-out items (0.00) are kept: they are why a total is lower
than what was ordered.
**Uber prints pick-up and delivery addresses on every receipt** and they were
thrown away. Captured as `route` [{label, time, address}], de-duplicated
because the template renders the whole block twice for narrow screens. Wording
is kept as printed ("Pick-up" on some receipts, "Pickup" on others) rather than
normalised, so a template change stays visible. This is the same block a *trip*
receipt uses for start and destination — rides are not ingested today, but the
reader will not need changing when they are.
Also stores source_email_subject/from, which order ingestion had left null on
columns that already existed.
Verified against the captured corpus: route on all 6 Uber fixtures, 5/5 items
on the GLOMARK grocery receipt including the sold-out one. Production data
updated by smarthome:docker/scripts/order-presentation-2026-07-27.sql
(81 descriptions, `backfill` tag, re-run clean). `route` and Uber line items
are parsed from here on only — recovering them for already-ingested orders
means re-reading the mail, which I7 idempotency refuses by design.
Both were found by looking at the data after the live backfill, not by the
suite — 105 tests were green while 85 rows were invisible and 4 were double
counted.
owner_id was NULL on every ingested order. Analytics scope on
COALESCE(t.owner_id, s.owner_id), and an ingested order carries no statement,
so the coalesce resolved to NULL and matched no owner. The rows existed in
`transactions` and appeared in no view in the app. Ingestion now sets
DEFAULT_OWNER_ID, and a regression test asserts the row survives the same
COALESCE scoping the UI uses.
[Family] orders are card-settled, not credits-funded. Their receipts name the
payer ("Payments Siddharth LKR 3,783.20") and no instrument, which an earlier
version read as credits. The card statement carries all four of them (CBA
...3893, exact foreign_currency_amount matches), so creating a transaction
duplicated spend already recorded — the double-count I5 exists to prevent.
They now record provenance only; the statement line is the transaction and is
what carries the `family` tag that keeps them out of budgets.
Production data corrected separately by
smarthome:docker/scripts/fix-order-backfill-2026-07-27.sql.
Also: reconciliation tests no longer assert global row counts.
reconcilePendingOrders() scans every pending row, so leftovers from other
files moved the totals — the source of an intermittent failure that only
appeared on the first run after a source edit.
Backfill dry-run over 130 real messages surfaced one 422: 'payments sum to 1.17
but receipt states 16.50'. The receipt is a mixed Uber payment —
Uber Cash $1.17 + Westpac ••••8032 $15.33 — and the card regex only matched
Visa|MasterCard|American Express|Amex, so an issuer-named leg was dropped
entirely. validateOrderTotals correctly refused it rather than recording $1.17
as the cost of a $16.50 order.
Anchors on the ••••NNNN masking instead, which also covers the form already
seen in the corpus ('Mastercard ••••3893 (CBA Ultimate) CHF 51.23'). Fixture
and regression test added.
Also repoints .env.test at the current postgres-personal container IP and
documents why: the container publishes no host port, so the address changes on
every recreate and the whole integration suite fails with connection errors
until it is refreshed.
None of these were caught by 105 green tests, because the code they live in was
barely tested and the HTTP path was not tested at all.
1. reconcilePendingOrders hardcoded category 'dining', so any order resolved
through the deferred path booked as dining regardless of merchant — a
Woolworths grocery order that parks and later reconciles was misfiled.
That reintroduced, through the back door, exactly the misfiling
resolveCategory() exists to prevent. Now calls it.
2. reconcileCardLeg never marked a statement line as consumed, so two orders on
the same card inside the +/-4 day window both bound to the same charge and
each booked its own credits remainder — double-counting spend. At 10-15
orders a month on one card that is not a corner case. Migration 0020 adds
matched_transaction_id with a unique index; the matcher now excludes lines
already claimed.
3. The ingest API returned HTTP 200 for every parse failure, and the Slack
alert fires only on non-200. So the single most likely production failure —
a provider template change breaking every order at once — was completely
silent. Split into NotAReceiptError (promotions, delivery updates, refund
and adjustment notices: 200, silent, expected traffic) and OrderParseError
(it IS a receipt and would not parse: 422, alerts).
Also: order_reference now anchors on Uber's own tripReference cell rather than
'first UUID in the document'. I had claimed to verify that the first UUID was
always the order UUID; that check compared against zero samples and was
vacuous. tripReference is present in all 29 captured receipts and, for ue-00,
equals the UUID the PDF redirect resolves to. The positional fallback remains
but only flags when there is genuine ambiguity.
Adds the API route's first tests — auth gate and error taxonomy — plus
anchoring regressions. 63 unit + 53 integration green on five consecutive runs;
corpus holds at 63/65.
Closes the three gaps left after the parser rebuild.
Refund amendments. ue-05 is a real refund: 'Previous total $49.94 / Refund
-$4.21 / New Total $45.73'. Uber reuses the order UUID across the receipt and
the amendment, so the two can be matched. The transaction is reduced in place
rather than offset with a second row — the order is one event whose cost
changed, and a compensating row would misreport both the meal count and the
merchant's spend. When the original was never ingested, nothing is invented.
[Family] orders now import instead of parking. Their payment line names the
payer, not an instrument ('Payments Siddharth LKR 3,783.20'), so no split is
recoverable and there is no card leg to reconcile against — they would have sat
pending forever, which fails the actual requirement to import and tag them.
Treated as credits, flagged as an assumption. Safe because the family tag
removes them from every budget regardless of instrument, and the LKR amount is
preserved with amount_aud left NULL rather than asserting an FX rate.
Ingest API. n8n now POSTs each message to /api/orders/ingest instead of parsing
in a Code node — the n8n sandbox has no require or fs, so a parser there cannot
be tested against the fixture corpus, which is the one thing that makes this
parser trustworthy. Auth is a shared secret, since machine callers have no
Traefik session header. Rejections return 422 and record nothing.
60 unit + 45 integration green on three consecutive runs; 63/65 corpus holds.
Ingestion now runs on the rebuilt parser. Three substantive changes.
Deferred card reconciliation. A 'MasterCard 8032 and/or credits' receipt never
states the split, but the card leg lands on the statement — Subway's $29.08
order shows $13.06 on 8032, so $16.02 was credits. For a live order that
statement is weeks away, so the split cannot be settled at ingest time. Such
orders are now parked with provenance and no transaction, and
reconcilePendingOrders() resolves them once the statement arrives. Backfill
takes the same path and resolves immediately. Migration 0019 adds the columns
that make an order resumable; applied to personal_test only, prod untouched.
Payment detection bug, found by the new tests: the old regex delimited the
'Paid with' line on a double space, which whitespace collapsing removes. Every
card and mixed receipt fell through to the credits branch — the Woolworths
receipt booked $60.93 of credits spend that never happened.
Category resolution reversed deliberately. Correction 1 said never default to
dining; the implementation of that sent everything unrecognised to 'other', and
knowing six merchants meant Carl's Jr, Taco Bell, Chilli India, Oporto, Schnitz
and Souvlaki GR all landed there. Grocers are an enumerable set and restaurants
are not, so match groceries explicitly and let the residual be dining.
Tests rebuilt on real captured receipts; the synthetic fixtures are deleted.
60 unit + 41 integration green on three consecutive runs.
The previous parser was written against synthetic fixtures shaped to match the
code. It invented a table layout DoorDash does not send, generated
order_reference from Math.random(), read the order date from a 'Date:' string
present in no real message, and detected [Family] by searching the body for the
substring 'family'. Its tests passed because the fixtures were built to satisfy
it. Against 36 real DoorDash and 29 real Uber Eats receipts it does not work.
Rebuilt from the real corpus. 63 of 65 now parse and validate; the 2 rejected
are correctly rejected — one is an order-adjustment notice and one a refund,
neither of which is a receipt.
Corrects an inherited diagnosis: the Mad Mex 'Discounts -$24.09' was recorded
as an HTML-flattening artefact masking a 'true discount of $9.45'. Parsing the
table cells structurally returns the same figures and no $9.45 exists anywhere
in the message — DoorDash genuinely prints a Discounts line that equals subtotal
plus service fee, and the components fail to reconcile on 32 of 36 receipts. So
the breakdown is stored as provenance and never gated on; validation instead
cross-checks the two independently stated totals and the payment line, which is
the number that becomes money.
Real-world cases the corpus forced, none of which were in the spec: [Family]
orders are LKR purchases for family in Sri Lanka (reading them as dollars
inflates ~200x), Swiss orders arrive in CHF, grocery 'Final receipt' mails carry
no Total Charged row, and a declined payment is printed alongside the successful
retry and must be skipped or it records money that never moved.
order_reference now comes from the Uber order UUID embedded in the body, or the
provider message id where DoorDash supplies no order id at all — never random,
so re-ingestion is genuinely idempotent.
Tests 12 & 13 asserted that every 'ShopBack Gift Cards' row becomes a transfer.
Resolved against the ShopBack purchase emails, 3 of the 14 matching rows are
Airbnb, 1 Shell, 1 Amazon — the bank descriptor's trailing token is a sequence
counter, not a brand code, so the description cannot identify what was bought.
Only $313.66 of $3,411.16 was ever reclassifiable.
Withdrawn rather than narrowed: making it safe needs ShopBack purchase-email
ingestion, brand resolution and an approval gate, to correctly handle 2
transactions in 20 months. Those two rows get handled by hand.
The integration suite shares one personal_test database and helpers.resetDB()
TRUNCATEs it with CASCADE, which reaches expense_metadata via the transactions
FK. With file parallelism on, queries.test.ts and participants.test.ts were
truncating rows out from under order-ingestion.test.ts mid-test, so a different
set of assertions failed on every run — including I6 (credits), I7
(idempotency) and I11 ([Family]), the three invariants the suite exists to
prove. Serialise the files.
Suite was reported 31/31 green; observed 29/31 then 28/31 on consecutive runs.
Now 31/31 on three consecutive runs.
One-off analysis, nothing built. Realistic baseline $4,140/mo -> $24,800 for
six months, against $89,770 already accessible ($81,017 loan redraw + $8,753
offset).
Records four corrections the raw data needs before any restatement:
misfiled Raiz/Vanguard/moomoo debits counted as spend, `other` credits read as
negative spend, `government` conflating ATO with rates/rego, and `fees` being
mostly annual.
CLAUDE.md gains two traps found while doing it: partial split coverage inside a
category is usually correct rather than a gap (only shared utilities and
subscriptions are split), and the loan repayment is voluntarily above contracted
($2,500 vs $1,190.54 per fortnight) with the difference recoverable via redraw.
CLAUDE.md gains the traps a new session would otherwise re-discover:
- Rules: a zero-condition rule matches everything (rule 43 would split all ~3,700
transactions); preview-then-apply-by-id is the safe pattern and why it beats
auto-applying on ingestion; how run provenance works.
- Shared expenses: transaction_splits.settled is dead data; getParticipantBalances
is correct and must not be 'fixed'; settlement cannot be attributed per trip.
- The shared loan: separate ledger, fixed 50% with a tracked receivable, why the
share must not be derived from actual payments, and why interest stays as spend.
- Extraction: balance assertions are the check that works, do not derive
opening_balance or add a totals assertion (both would be tautological), Gemini
invents summary fields it was not given, empty statements must not throw, FX is
per-date, and CSV comparisons need millisecond ordering.
The design doc records Phase 0 as done - including that the original Phase 0 plan
was wrong, since reading the code first is what prevented breaking a working
balance page.
Known Gaps lists what is open: the unbuilt phases, 11 failing assertions, the
uncategorised Up rows, and the CSVs sitting in 030490e's history.
The trip view showed Total Owed / Settled / Unsettled per participant, with the
last two derived from transaction_splits.settled. Nothing sets that flag - its
only writer was /api/splits/settle, which no UI calls - so it is false on all 673
splits and every trip reported 100% unsettled, including trips already paid in
full. Molina has paid $20,782.79 against $19,556.07 of splits and the Europe trip
still showed her entire share outstanding.
A correct per-trip figure is not computable either: split_payments records only
from, to, amount and date, so a payment cannot be attributed to a trip. The trip
view now shows each participant's share and points at Shared for what is actually
owed, which is where settlement genuinely lives.
Also removes /api/splits/settle. It was unreachable from the UI but live on its
URL, and a single call with participant_id would mark every one of that person's
splits settled - writing a flag nothing reads. Settlement will be reintroduced
against settlement contexts (docs/shared-expenses-design.md).
getParticipantBalances is deliberately untouched: it computes splits minus
payments, which is coherent. Excluding settled splits there while still
subtracting the payments that settled them would double-count.
The loan is a separate ledger, not a settlement context: a contribution must
never be able to settle a dinner.
The share is fixed at 50%, not derived from actual payments. During Sonu's leave
the obligation did not change, only the payment did - a percentage-of-actual
model would silently redefine her share as 30% and make the shortfall vanish. So
the model needs an expected schedule alongside actual contributions, with the
difference as a tracked receivable. Currently $4,000.00 over Jul 2025 - Jun 2026.
On interest: recorded the mechanics (it is debited to the loan and repaid as part
of the balance - the reconciliation is exact) alongside the counter-argument that
$16,523.64 left and bought nothing, which is what an expense is. Recommends
keeping it as spend with a fixed/discretionary grouping to address the real
concern, but flags it as a judgement call rather than settling it.
Three problems that look separate are one: the app records money moving, and
separately records who owes whom, and the two never meet.
Documents what is broken with evidence - two half-built settlement models,
settlements existing twice unlinked, and Sonu's $37,980 of loan contributions
sitting unrecognised as generic transfers - then proposes settlement contexts,
payments as transactions rather than a side table, and loan co-ownership.
Nothing built. Five open questions, two of which are decisions about the
arrangement rather than the software.
Apply History listed only counts - '13 matches · 13 transactions' - which reads
identically whether the run renamed a merchant or split every transaction with
another participant. Revert is destructive, so that is not enough to decide on.
Two additions. Migration 0017 records rule_id, rule_name and source on each run:
rule_name is denormalised so history stays readable after a rule is edited or
deleted, and there is no FK so deleting a rule cannot cascade away the audit
trail. Both write paths now populate it - the condition-matched run and the
selection-based quick action.
And rows expand to show the run's snapshot set against current values: which
transactions were touched and what changed on each. Rows changed by something
else since the run are called out, because reverting restores the pre-run value
and would discard that later edit.
Runs recorded before this show 'Unknown rule' - the rule they came from is not
recoverable.
Selecting a rule now shows the transactions it would alter, so a subset can be
ticked and applied rather than trusting a bulk run. The apply step takes explicit
transaction ids (the existing bulk apply_rule path), so what you tick is exactly
what changes - a rule whose conditions are too broad cannot reach further than
the preview showed.
Matches are split into 'would change' and 'already correct'. A merchant
normalisation rule matching 400 rows where 380 already hold the right value is 20
changes and 380 rows of noise; only the 20 are listed.
Preview is offered for every rule including manual_only quick actions, which
previously had no way to see their reach at all. A rule with no conditions
matches every transaction - that is how apply already behaves, so the preview
reports it prominently rather than hiding it.
Applies still snapshot to rule_apply_runs, so they remain revertable.
A git add -A in 030490e swept the Wise CSV exports into the repo and they were
pushed. They contain account numbers, an IBAN, payer and payee names and full
transaction detail.
This removes them from HEAD and ignores the directory. It does NOT purge them
from history — 030490e still contains them.
getPendingReconciliations treated every unreconciled manual transaction as
awaiting a matching statement row. Cash never appears on a statement, so a cash
entry sat in the queue indefinitely being offered matches within 3 days and 1%
on amount - and accepting one is silently destructive: reconciled manual rows
are filtered out of every query, so the cash spend disappears while the card
transaction it matched claims to be that same spend.
Migration 0016 adds transactions.payment_method (card | cash | bank_transfer |
other, NULL = unknown) with a CHECK constraint and a partial index. The notCash()
fragment excludes cash from both halves of the reconciliation query - the pending
list and the candidate match subquery, which aliases the manual row as m.
Only cash is excluded. Bank transfers do appear on a statement now that
transaction accounts are imported, and NULL means unknown, so both stay
candidates and every pre-existing row behaves exactly as before.
ATM withdrawals deliberately stay categorised as spend rather than transfers.
Treating them as transfers is only correct if every cash purchase is logged;
with partial logging it silently deletes the unlogged remainder from spend.
Four related fixes to spend correctness.
Balance assertions. Nothing checked that a statement's transactions add up to
its closing balance. getStatements now computes opening + movement - closing
and the statements page flags any statement that does not reconcile. Sign
depends on what the balance means: on a credit card or loan it is what you owe,
so spending increases it; on a transaction or offset account it is what you
hold. 11 statements currently fail, $4,177 unexplained - including two adjacent
ANZ statements off by exactly +/-$230.38, a transaction filed against the wrong
one.
Category normalisation (migration 0015). Categories arrive from Gemini (which
writes straight to Postgres from N8N), CSV import and manual edits, so the rule
belongs in the database - same reasoning as normalize_statement_type in 0013.
Adds normalize_category(), triggers on transactions and transaction_overrides,
and CHECK constraints. Backfilled 122 rows: 19 'payment' ($42,569) to transfers,
14 'refund' recovered to the merchant's usual category, and title-case duplicates
folded into their canonical spelling - 'Shopping' and 'shopping' had been
counted as separate categories by every GROUP BY.
Refund netting. Monthly analytics counted only debits, so a refund was counted
nowhere: excluded from spend by type, and not income by category. A $2,888.92
Expedia purchase refunded in full eight days later still read as $2,888.92 of
spend. SPEND_SIGNED and NET_SPEND_ROWS bring refunds in as negatives; 'income'
joins the excluded categories so incoming money cannot leak in as negative
spend. $36,773 across 93 rows now nets correctly.
Drill-down share. The insights drill-down used my_share_percent ?? 100, which
ignored transaction_splits entirely and repeated the bug fixed in ab00f8c one
layer up. getTransactions now returns my_share_pct and my_amount resolved
server-side, and the table shows gross alongside your share.
The split-adjusted spend expression assumed a transaction with no split row for
me was entirely mine. That is wrong when a transaction is allocated fully to
someone else: I paid, they owe all of it, and there is no row for me to match.
Both branches of the CASE missed and the ELSE charged me the full amount.
24 transactions were affected, all travel bookings between 2026-01-09 and
2026-06-26, overstating my spend by $8,579.07 across every analytics view.
Adds myShare/mySplitOf to analytics-sql.ts, which fall back to
100 - (sum of everyone else's shares) instead of 100, and applies them to all
five analytics routes. Centralised for the same reason as EXCLUDE_NON_SPEND:
the expression was duplicated five times and had already drifted.
The transactions page rendered the statement's native amount through a
formatter hardcoded to AUD, so a USD row displayed its USD figure labelled as
dollars while every analytics query counted the converted amount_aud. Same
transaction, two different numbers depending on the page.
getTransactions now returns the statement currency, the amount column shows
amount_aud with the native figure beneath it when the two differ, and the split
and duplicate modals seed from the converted amount (a duplicate becomes a
manual AUD row, so the native figure would be wrong there).
Settlement balances had the same split: getParticipantBalances and the
per-participant balance route summed raw amount while trip totals summed
amount_aud, so a shared foreign expense would net a USD figure against AUD
ones. All three now agree on amount_aud. No change to current balances - every
statement in the database is AUD today - but correct once Wise data lands.
Regression from the analytics scoping work. Chain:
1. Three transactions have a NULL category. The old WHERE clause
`COALESCE(o.category_override, t.category) NOT IN ('transfers','investment')`
evaluated to NULL for those rows, so they were silently dropped — which is
the bug EXCLUDE_NON_SPEND fixed by defaulting to 'other'.
2. Including them exposed that the SELECT still used the non-null-safe
COALESCE, so MODE() returned NULL for the affected merchant
("Shared expenses carryover (SplitMyExpenses)", a manual transaction that
the LEFT JOIN fix also newly included).
3. formatCategory(null) threw "Cannot read properties of null (reading
'split')" and took down the whole page render.
Fixed at both layers:
- Both merchants routes now use the EFFECTIVE_CATEGORY fragment, so the API
cannot emit a null category.
- formatCategory tolerates null/undefined and returns "Uncategorised". An
uncategorised transaction should never be able to crash a page.
Verified against the live DB: 200 merchants, zero null categories, and
/merchants /transactions /insights /budget /statements all render 200.
A loan repayment is not an expense. A $3,000 mortgage repayment is roughly
$1,200 of principal (equity — a balance-sheet move) and $1,800 of interest (the
only part that is genuinely spend).
Migration 0014:
- transactions.principal_amount / interest_amount, populated only when the lender
itemises the split on the repayment row
- statements.interest_rate, scheduled_repayment, repayment_frequency,
redraw_available, loan_term_months
- normalize_repayment_frequency() + trigger, so "Fortnightly", "Bi-Weekly" and
"Every 2 weeks" all land on 'fortnightly'
Two statement shapes are handled. Where the loan statement lists repayments and
"Interest Charged" as separate rows (the common Australian case), transaction_type
already does the work. Where a lender itemises the split on the repayment row,
that row is typed 'payment' and would be skipped entirely — losing the interest.
New SPEND_ROWS / SPEND_BASE fragments in analytics-sql.ts count such rows at
interest_amount instead of amount.
Adds the loan_interest category (+ colour, and the missing fees colour).
Verified against the live DB with a synthetic ANZ home loan statement: a $3,000
itemised repayment plus a $10 service fee moved April spend by exactly $1,810,
with the $1,200 principal excluded and still retained on the row. Test data
removed and the figure confirmed back at its original value.
Groundwork for importing bank and loan statements alongside credit cards.
statement_type was whatever free text Gemini put in account_type ('Credit Card',
'credit card', 'credit_card', 'Business Card', 'ACCESS ADVANTAGE',
'multi-currency account'). The UI coped only by doing .includes("card"), which
breaks as soon as bank and loan statements arrive.
- Migration 0013 adds normalize_statement_type() + a BEFORE INSERT/UPDATE
trigger and a CHECK constraint over credit_card|transaction|savings|loan|
offset|investment|other. The trigger means the N8N workflow keeps working
unchanged while it still sends free text. Raw value stays in account_type.
Backfilled 99 existing rows.
- src/lib/statement-types.ts mirrors the vocabulary for the UI; statements page
now filters by the real types and headlines balance vs amount due per type.
Analytics were scoped with INNER JOIN statements + s.owner_id, which silently
dropped all 180 manual/CSV transactions (statement_id IS NULL) from every
report. Switched all six routes to LEFT JOIN + COALESCE(t.owner_id, s.owner_id)
via shared fragments in src/lib/analytics-sql.ts, so the transfers/investment
exclusion that stops card-payment double counting stays consistent. Also
extended that exclusion to trip analytics, which had none.
Drive-by: /api/analytics/subscriptions was returning 500 on an unserialisable
BigInt from COUNT(*) + 1.
Verified against the live DB: monthly spend picks up the previously invisible
manual transactions (Apr 9,849.91 -> 14,286.70) and all four analytics
endpoints return 200.
A rule flagged manual_only never runs in the apply-all pass; instead it shows
as a button in the transactions bulk bar and applies its actions to the current
selection (conditions ignored — the selection is the condition). Recorded as a
rule_apply_run, so it reverts from Rules -> Apply History like any other run.
Motivation: tagging Home + splitting 50/50 with Sonu was two bulk actions every
time. Now it is one click, and any other combo can be defined the same way.
Extracts the action-application and snapshot logic from the apply route into
src/lib/rule-actions.ts so both callers share one implementation — splits upsert
rather than delete+reinsert, so settled flags survive.
siddharthd