Bills are LINE ITEMS not documents: group by INVOICE_NO across three GST
eras (GST pseudo-line / 2026 columns-only / pre-GST zero-tax). 46 raw
ticket statuses -> 5 via ordered pattern map; OLD_ID re-key chains
collapse so superseded rows do not double-import. YES/NO flags respect
real vocabulary (GENERAL ENQUIRY != yes); ATM cell value becomes the
module provider. SID 329 test client + orphan refs skip (counted notes,
distinct from commit-blocking problems). District normalize, NO DATA
scrub, payment normalize, richer-row merge for duplicate service rows,
shared-number-space series seed. Dry run over the real export: 273
clients, 150 documents = Rs 33,16,800.24, 72 payments, 1,826 tickets,
667 interactions, INVOICE seed 10252 (matches APEX next_val). Suite 358
green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
parseWorkbookCsvs + stageApexFull (pure validation pass, zero writes) +
commitApexFull (one transaction): clients w/ encrypted DB passwords +
secretary contacts, YES-flags -> live client_module rows for the 5
modules, branches, bill_filed -> QT/PI/INV history with paisa-exact
CGST/SGST split + payment/allocation rows + series seed, 4.5k tickets
with status/date mapping, SMS/RTGS service data -> client_module
provider/username/password_enc/details, project lists -> interaction
notes with tolerant APEX date parsing. Plaintext passwords never stored
or audited; missing HQ_SECRET_KEY is a blocking problem. CLI:
npm run import:full -- --dir <folder> [--commit]. 7 new tests; suite
359 green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Thomas Joise