What we shipped, in plain language.

The upload took the first sheet in the workbook, blindly. One real file carried fourteen sheets and position one held a TA sheet for a different, earlier order — which parsed perfectly and named a project that was not the one being uploaded. Sheets are now scored on whether they name a project, carry a planned start date and hold task rows; the best one wins, every task sheet found is offered, and an explicit choice still overrides. A workbook with no task sheet at all is refused by name instead of falling back to sheet one.

The header block was read only below the column headings, only from rows carrying a particular flag, and only from the end-date column. Real files do none of those three reliably, so the whole grid is scanned now, in any column. Labels are matched exactly, then as a phrase inside a longer cell, then by close spelling — and a column heading can never be mistaken for a header label, which is what previously turned the heading row into data and the first task into a heading.

The first heading row is not always the complete one: in that file only the second row named Employee ID and Employee Name, so every task imported unassigned. Columns the first row omits are now filled in from a later one, and columns it does name are never overridden.

When a value really is missing, the message now names what the pass did read and lists the cells it could not match, with their references — so a sheet the parser does not recognise stops looking like a blank cell.

The assignee name reached for a field the list query did not fetch, so every row paid a second database round trip — and when that re-fetch found nothing, the whole list raised an error instead of rendering. Serving a 78-task project now costs four queries.

Department is now a visible column on the task worksheet, and every other field the task list can answer is added alongside it — hidden by default and reachable from the column chooser, so the default sheet stays the width it was.

What that adds: assignment (department, role), the dates the plan actually moved on (revised end, actual start and end), ordering and verification, the scheduling pass’s own output (earliest and latest start and finish, total and free float, weight) and provenance (WBS element, task definition, generation date). The new columns are deliberately read-only: a column that looks editable but is dropped on save is worse than one that plainly is not.

One vendor held a ₹5,00,000 approved advance with nothing allocated, and two open invoices. The manual payment screen showed the invoices and had no concept of a credit at all.

Open credit is now reported on the screen, in two tiers split on currency. A credit in the same currency as a selected invoice could have cleared it, so it is named and totalled and the payment waits on a written acknowledgement — a tick box reading “pay without applying ₹X of open credit”. A credit in any other currency could not have been applied to this payment anyway; it is reported and blocks nothing, because asking for a confirmation on a payment that was never at risk is how confirmations become reflexive.

The rule lives in the endpoint, not only on the screen. It is deliberately soft: an advance is not proof that it belongs to the invoices being paid, and a hard refusal would strand an operator who has no right to apply the credit either. A bypass is stamped on the payment’s own notes naming each credit left unapplied — and nothing is stamped when the box was ticked but no credit was actually open, because that would be a false statement on the document.

An invoice showing ₹10,32,500 credited and ₹10,15,000 outstanding with no payment against it was explaining nothing: the ₹17,500 difference was TDS withheld at posting, and the whole withheld amount had been pushed into the allocated column because that was the only lever that made the outstanding figure come out right.

So the sub-ledger reported an allocation that never happened and could not say why an unpaid invoice was part-closed. The general ledger was never wrong.

TDS is now stored as TDS, the allocated column goes back to meaning what payments actually cleared, and the Vendor Ledger grid has a TDS column — without which filling in the field would still have shown the operator nothing. Historical rows compute the same outstanding figure either way.

Administrators: run backfill_vendor_ledger_tds --tenant <id> to restate existing invoices. It reports by default and writes only with --apply. No balance changes — an amount only stops being called an allocation and starts being called TDS.

Both AR invoice screens now show the customer’s open advances and offer to adjust them. They prompt; they never knock an advance off automatically — which invoice an advance belongs to is a judgement, not a lookup.

Administrators: after upgrading, run batch_migrate --app fi_ar and setup_rbac, then configure a GST advance account per company. Issuing a voucher refuses by name without a tax-on-advances account.

Tax on an advance is due in the month the money lands, reported in GSTR-1 Table 11A, and reversed at Table 11B when the tax invoice finally goes out. There was no document for it.

The voucher is an overlay on the customer receipt, not a second cash document. The receipt already owns the bank leg, the ledger, the reconciliation and the reversal engine, and a parallel document would fork all of them. A receipt already applied to invoices is refused by name — that supply is taxed on its invoice — as is a foreign-currency receipt, because GST is a rupee liability and nothing here gets to pick the conversion rate.

The amount received is always tax-inclusive, so the tax is the subtraction of taxable value from gross, never a rate applied forward. Computed forward off a rounded taxable value, the columns fail to add up by a paisa on roughly half of all amounts — and a statutory document whose own columns do not add up is not defensible.

The tax posts as its own journal, so a receipt with no voucher posts exactly the two legs it always did. Adjustment resolves the invoice rather than trusting the reference it was handed: an id naming nothing, another customer’s invoice or a draft is a bad filed return.

Where the rate cannot be determined the law says 18%; where the place of supply cannot be determined it says treat the supply as inter-State. Both are applied — and both are recorded as fallbacks on the voucher, because a default nobody can see is indistinguishable from a decision, and whoever prepares GSTR-1 has to know which Table 11A rows to revisit before filing. A tenant-configured default tax code is not flagged: the tenant chose it.

Rule 49 caps the voucher serial at sixteen characters, so an over-long numbering pattern is refused rather than truncated into a collision.

A help-documentation seed named a field that lives on a different record, and the chart-of-accounts seed checked for an existing chart by its filing type — which misses a tenant whose chart is filed under an older type code, and then collided with the accounts that tenant already owned.

The chart seed now works out every code it would write and stops if any of them already exists, which is what it always claimed to do. Neither change touches an existing row.

The employment application and job request forms take their second language from the applicant’s mother tongue. Gujarati was selectable as a mother tongue but held no labels, and an unheld language falls back to the default — so a Gujarati applicant was handed the Tamil form and nothing reported it.

The form embeds the typeface it asks for, so the Gujarati face ships with it. Without that the form prints correctly on a developer’s machine and renders every second line blank on the print server. A test now asserts that every language in the catalogue has its face on disk, rather than leaving that to be discovered on paper.

The adjustment entry that carries the difference between what the timers logged and the total a supervisor states now stores whether it adds or removes time, rather than leaving that to be inferred from the figures around it.

A new scheduled report sends each person their own timesheet for the day: the tasks they worked, the hours booked against each, and what remains open. It is a real schedule like any other report — it can be re-timed, paused or removed.

Its audience comes from the data, not the schedule: whoever booked time that day is who receives a mail, so nobody has to be added to a recipient list when they join and nobody keeps receiving an empty sheet when they leave.

The delivery detail page switches to the colour and size matrix as soon as any item on the note has a variant, and then sends every line to it. A line with no variant matched no cell: it rendered nowhere, could not be edited, and was absent from the grand total — while the header, which sums every item, still counted it. One note read 7,738 in the header and 2,515 in the grid.

Items are now placed through a map that consumes each item exactly once. Whatever cannot be placed — no variant, a missing row or column attribute, or a second line landing on a cell another line already holds — is collected and shown below the grid in a flagged, editable table with its own subtotal, plus a combined product total.

The page also gained a Lines total chip summing every item including unsaved edits, and shows an amber Header says… chip when the two disagree, so a mismatch is visible rather than silent.

A marker is now linked to the production orders it is being cut for, with each order’s own quantities visible while the marker is planned. The marker form and list were rebuilt around that link, and production planning answers the order and quantity lookups the form needs.