The bars were capped at 36px but centred in half-month slots, so on a
wide screen a month's pair drifted ~65px apart. A fixed 32px bar with a
3px gap keeps them together (recharts still shrinks them if space runs
out).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Text PDFs already get it from pdftotext; photos and scans have no text,
so 4 of 5 undated POs were photos that do print a PO Creation Date.
- n8n PO reader (stored patch): asked for order_date = 'PO Creation
Date' (day/month/year), never Required By or Printed; the callback
passes it on.
- process-batch accepts extracted.order_date (YYYY-MM-DD); the PDF text
wins when both exist.
- create_purchase_order / update_purchase_order / move_to_purchase_orders
take order_date, so agents record it too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The reader always read the slip date (and receiver, site name, customer
VAT, supplier registration), but dn_batch_notes never had those fields,
so they were dropped on save and every receipt was dated the moment it
was posted. The Financial dashboard dated orders by upload date, so 24
of 41 POs (R7.08M) sat in no month and deliveries piled up in the month
they were booked: 'R1.1M ordered, R2.4M delivered'.
- scripts/pb-add-dn-dates.py: adds the five header fields,
delivery_notes.posted_at and purchase_orders.order_date (applied).
- scripts/dn-backfill-dates.py: filled 270 documents from the stored
readings, moved 192 receipts to their slip date (booking time kept in
posted_at) and read 36 PO order dates from the PDFs (applied; replaced
values in docs/audits/2026-10-11-dates-backfill-backup.json).
- Posting refuses a slip without a delivery date instead of using today;
receipts record posted_at. The review page shows the date and locks
Post with a plain message when it is missing.
- PO intake reads 'PO Creation Date' from the PDF text (pdftotext).
- Dashboard tile and monthly chart date orders by order_date only; an
order without one is undated, never put on its upload month.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Orders and deliveries now share a grouped bar chart. Orders with no date
(most of The Hills' value: date_uploaded is empty and the `created`
fallback does not exist on purchase_orders) cannot be placed on a month,
so the chart now totals them in a note instead of dropping them silently.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Row ids with the hours_worked values from before the 2026-10-09
paid-hours recalculation, so the run can be reversed.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
The "Received this month" tile linked to the Suppliers page, which
always opened on "All", so the page showed a different number than the
tile. Add a validated parseInitialPeriod helper, pass the parsed
period through the page as initialPeriod, and have the tile link carry
?period=thisMonth.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The "Received this month" sub-line always said "suppliers", so a single
supplier read as "Owed to 1 suppliers". Use the singular when the count
is 1.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
date_uploaded and created are UTC timestamps written with
toISOString(), but monthKey is the South African month. Comparing the
raw YYYY-MM prefix put an order uploaded at 00:30 SAST on 1 November
into October. Resolve the SAST work date first, then compare months,
and skip values that cannot be parsed.
Also reword two workforce comments that referred to a worker as "he".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Intl renders en-ZA currency and numbers with U+00A0, so the tile helpers
normalise it before asserting. Written as a literal that byte is
invisible in the source and easy to "fix" into a plain space by accident.
An explicit escape says what it means.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The overview test asserted the previous all-time sub-line wording and
that employee_rates was read with a site filter, both of which the cost
workforce commits changed on purpose.
Pins the clock so "this month" is a fixed window, and gives the unpriced
worker an October entry so the new month-scoped warning is actually
exercised rather than skipped. Added a case asserting rate history is
read unfiltered, which is what lets a worker who moved sites keep the
rate he was paid at.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
"POs (total)" repeated the number already shown under Open PO value and
said nothing about deliveries, so the page had no figure for what came in
this month or what that leaves owed to suppliers.
Replaced with "Received this month", computed with the shared
deliveryLineValues helper -- the same series behind the Suppliers page's
"Payable this period" -- filtered to the current SAST month. Under the
app's rule that we only pay what we received this month (see
src/lib/supplier-period.ts), delivered-this-month and owed-this-month
are the same number, so one tile answers both and cannot drift from the
Suppliers page. The sub-line names how many suppliers that is; with none,
it says so rather than showing a bare R0. The tile links through to
Suppliers with the current site selection preserved.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PO spend summed total_value of anything uploaded this month, including
failed and still-processing placeholder rows -- money that was never
committed. It also ignored orders raised through the MCP create_purchase_order
tool, which carry no date_uploaded at all and so never appeared in any
month.
The tile now sums real orders only, matching the rule already used for
open value, and falls back to `created` when there is no upload date.
Month membership is a YYYY-MM prefix against the SAST month key. The
sub-line says how many orders that was, so a plausible-looking total can
be sanity-checked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
"Open PO value" summed total_value of every processed order ever and
never subtracted what had been delivered, so it grew without bound and
read as total committed rather than still owed. A R1 000 order with R400
delivered reported R1 000 open.
It now comes from computeSupplierLedger -- the same computation the
Suppliers page uses -- over the items of real in-scope POs and the
delivery lines belonging to them (by booking snapshot, or by parent PO
item for older unbooked receipts). Clamped per PO, so an over-delivered
order cannot net off another order's open balance and understate what is
still owed. The sub-line counts orders still actually open rather than
repeating the same number as a "total POs" figure elsewhere on the page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Suppliers page built its per-delivery-line value series inline to
feed the "Payable this period" KPI. Extracted it to deliveryLineValues
so the Financial landing can show the same figure for a single month
without re-deriving the rules in a second place, where they could
quietly diverge.
Behaviour is unchanged: booking snapshot net wins, else qty_received x
the PO line's rate; receipt documents only; lines against POs outside the
page's scope are dropped; lines with no date_received are skipped. The
returned rows now also carry the PO id so callers can attribute a total
to a supplier.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
fuel_logs is site-scoped, so it holds equipment fill-ups as well as
fleet ones, and the "Total litres pumped" / "Total fuel spend" tiles —
labelled as fleet — were summing both. The same diesel was counted here
and again on Plant.
Fleet-ness now comes from the log's EXPANDED vehicle (the query already
expands "vehicle") rather than from membership of the site-filtered
vehicles list, so a truck that has since moved to another site still
counts at the site where it was actually fuelled. Unclassified equipment
counts for neither bucket. The flagged count follows the same rule.
Extracted into a pure helper next to the page so it is testable without
standing up the PocketBase reads.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Timesheet entries are stamped with the site they were worked at and are
never joined to the worker's current home site, but both the cost loop
and employee_rates were filtered by employee.site. A worker who moved
from A to B therefore lost his A hours when viewing A (he was no longer
on A's roster) and never had them counted when viewing B — the hours
only reappeared under "All sites".
Hours and cost now iterate the in-scope timesheet entries and price each
employee through an unfiltered lookup, so they follow the site the work
happened at. An employee id missing from that map keeps its hours but
counts as unpriced. The roster stays site-filtered and still drives the
figures that are about who works here now: headcount, permanent/casual,
idle 30d, missing rates and the inactive filter on clock events.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The workforce KPI row showed total cost (all-time) and hours this week,
with nothing about the current month — the two figures a cost-to-date
conversation actually starts from.
Cost (this month) sums hours x resolveEmployeeRate over entries whose
date falls in the current SAST calendar month (YYYY-MM prefix, so
PocketBase's "2026-10-01 00:00:00.000Z" needs no parsing). It still
prices through the dated rate history, and its sub-line falls back to
the all-time total so the longer view is not lost. Hours (this month)
is the matching hours figure, with the week alongside it as sub-line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Set-aside delivery documents were kept but shown nowhere, so they looked
lost (e.g. The Waste Group slip T025511, set aside 'Waiting for
invoice'). The inbox now has a Set aside button: every set-aside
document with why (in plain words) and when, a link to its page, and a
Restore button that puts it back in the inbox through the same restore
the agents use, recorded with who and why.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The automatic refresh already handles a PO that moved on; when the
banner still shows, only Try matching again helps.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Mon-Thu 07:00-16:30, Friday 07:00-15:00, last Friday of the month
07:00-12:00 (weekends use the Mon-Thu bands). Each shift end has a
10-minute band either side; arrivals up to 07:10 pay from 07:00.
Lunch now phases in from 5h to 5h30 instead of a 30-minute cliff, so
a normal day pays 9 / 7.5 / 5 hours.
The worker calendar judges each day against that day's expected hours,
so a normal Friday is green, and the worker page shows the schedule.
The recalc script recognises rows already rewritten by the first rule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A scan in up to 07:10 pays from 07:00; a scan out between 16:20 and
16:40 pays to 16:30; a session over five hours loses a 30-minute lunch.
A normal day is therefore exactly 9 hours (green on the calendar and
matrix); late arrivals, early exits and overtime use the real times.
Applied on clock-out, in missed-scan corrections and their preview.
The calendar keeps the raw scan times and labels hours "Paid hours".
Includes a dry-run-first script to recalculate existing QR entries; it
only rewrites rows still equal to their raw scan total.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- inbox_board: while a site question is open, blocking lists only that
question (the reader's older doubts were misleading Foreman, e.g. the
corrected FH Chamberlain order). next now points agents at
prepare_delivery, and at the site manager for Ready-to-post slips;
agents no longer book.
- Foreman skill: describe parked documents only from the open question;
VERDICT counts only what is new this run (still-waiting items become a
count line); quote refusals and their hint instead of guessing a cause;
candidate_closed on regrouping means a document in the batch is set
aside or posted.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Foreman (MiniMax) works the delivery-note inbox: checks every slip
against its photo and PO, fixes misreads, groups kits, prepares clean
slips with prepare_delivery (Ready to post, never booked) and asks the
site about the rest. The review skill spot-checks his work, moves
answered site questions on to Ready to post and applies site-confirmed
receipt corrections. The site manager pulls the trigger.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Agents can now save a whole receiving plan (lines by number, kits, NIL
auto-close) and leave the slip in Ready to post; the site manager checks
the totals and presses Confirm delivery, which posts exactly what was
saved. Quantities over the PO need acceptExcess (site-approved) and a
slip with an open site question is refused.
Confirm delivery on the web now refreshes a slip whose match went stale
because another slip on the same PO was booked, re-applying the saved
decisions instead of sending the user back to an AI re-match.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Keeping the old value when the quantity or line changes would book, say,
162.5 m at the value of 65 m. Adds hints for both value refusals.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
