scripts/backup.mjs (npm run backup): pg_dump the DB to a timestamped compressed
pg_restore-able file under ./backups (gitignored), keep newest HQ_BACKUP_KEEP (14).
Locates pg_dump on PATH or the Windows PostgreSQL install dir. Verified against live
data: 304 KB dump, 32 tables, valid archive. Backups hold plaintext creds (D32) so
backups/ is gitignored + documented as access-controlled. Pending: schedule + off-box
copy + DB password out of server.ts.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Owner has no HQ_SECRET_KEY, so the D31 unlock-with-key path was impossible. crypto is
now fully keyless: encrypt/decrypt ignore keyHex (params kept so a key can be wired back
later); decrypt returns '' for any leftover legacy AES value (unreadable → treated as
unset, never fed back as ciphertext). Key-based /admin/migrate-plaintext replaced by
keyless /admin/clear-legacy-credentials (clearLegacyCiphertext) which NULLs unreadable
pre-D31 values. .env.example marks HQ_SECRET_KEY removed.
Ran the cleanup on live data: 76 SMS + 110 client DB passwords were unrecoverable and
cleared (usernames intact) — must be re-populated in plaintext (APEX re-import / re-entry).
Also fixes the one real defect the D31 adversarial review confirmed: maybeRefreshSmsDaily
still had a 'keyHex==="" -> no-op' guard that would silently kill the DAILY auto-refresh
once the key was gone (manual refresh masked it). Removed; the daily pull now runs keyless
like the manual path. Full suite 416 green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Removes credential encryption at rest so the SMS-gateway logins (and other portal/DB
passwords, module secrets, Gmail token) are usable daily without a held-out
HQ_SECRET_KEY. Keystone: crypto.encrypt() stores tagged plaintext ('plain:'),
decrypt() untags keylessly and still reads legacy AES *with* the key to migrate it.
Drops the six 'refuses/cannot without HQ_SECRET_KEY' guards. Scope is the reversible
credential vault ONLY — login passwords stay one-way scrypt-hashed (@sims/auth).
Secrets still kept out of list payloads + audit rows; reveal still managerial+audited.
migrateLegacyCiphertext + POST /admin/migrate-plaintext (owner) rewrite existing AES
rows as plaintext using the key one final time (idempotent, one audit summary), after
which the key is never needed. SECURITY: a DB dump now exposes these in the clear.
Tests updated to the plaintext policy; new migrate-plaintext.test locks decrypt +
migration + idempotency. Full suite 416 green; typecheck clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Pull each SMS client's actual credit balance from the provider panel instead of
manual entry. sms.balance_api_url (dated setting, default 164.52.195.161/API/
BalAlert.aspx) returns pipe-delimited code|STATUS|payload; parseBalanceResponse
reads the trailing number and flags ERROR lines, never guessing. Uses each
subscription's already-imported username/password_enc — the automated path decrypts
directly (no reveal-audit spam) and audits sms_balance ONLY when it changes;
per-poll telemetry lands in the unaudited sms_balance_check table (SQLite schema +
PG migration 010). Managerial 'Refresh from gateway' button (POST /reports/sms-
balances/refresh) + per-client refresh; scheduler does one pass per day
(maybeRefreshSmsDaily). Low-balance alert on the Dashboard + cockpit tile. Needs
HQ_SECRET_KEY (import-time key) to decrypt — no-ops with a clear message otherwise,
never stores plaintext. Parser unit-tested; full suite 408 green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Durable specs for tickets/reconciliation/MIS/sms-balance/tds/audit-viewer/
notifications/einvoice (+ renewals, delivered) with schema/repo/route/UI/test
detail and a collision-safe build order. WhatsApp design was transiently
classifier-blocked; to re-run.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- recordPayment locks the target invoice row (SELECT ... FOR UPDATE, Postgres
only — SQLite is serialized by the connection mutex) before reading
outstanding, so concurrent payments cannot over-allocate on prod.
- INDEX_DDL now runs AFTER migrate() in openDb, so indexes referencing
migrate-added columns (owner_id, lockout cols) build on older DBs.
- D24 remediation recorded in 06-DECISIONS; test counts 377 -> 389.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Each module declares its own operational fields (field_spec JSON);
per-client values organized by that module schema; generalized encrypted
secret map. Flexibility mandate made concrete: new module or new field =
no code, no migration. Queued after the D20 trial import.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Survey verdicts for all 25 real APEX tables; founder calls locked:
full ticket desk (4,567-ticket history), client_module service columns
(provider/username/password_enc/details), client_branch, five modules
= SMS, RTGS, WhatsApp, Mobile App, ATM.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Locked decisions: pg driver behind an async Db interface with dual
engines (sqlite for fast tests, pg for prod), bigint paise, TEXT ISO
dates, jsonb payloads, numbered migrations, ?->$n in the adapter,
DATABASE_URL selects the engine. Build order: async refactor first
(zero behavior change), then schema, adapter, dual-engine suite, smoke.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
STATUS: count + review summary, D18 sentence in the intro, honest migrate()
column list, routes table gains /reminders /documents /profile /import and the
collapse/user-menu shell description. README + CLAUDE: 345. D18 record in
06-DECISIONS gains the review outcome. Spec banner notes the post-review
deltas (CSV support columns, LIKE filters).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Records the decision the code already cites (staff_user IS the employee
table; rename would churn FKs and audit names) and the reviewed deviation
from the funnel spec's blanket pagination rule: GET /employees returns the
bounded console-user set whole with total, no page/pageSize.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The historical HQ-1/2/3a build plans and the live-preview spec referenced sibling
Store apps (apps/store-server, apps/backoffice) as copy-sources and cited Store-only
docs. Rewrote those to describe the patterns intrinsically and to name the Store
product/repo where the scope boundary is genuine, so each plan stands alone in the
HQ repo. Technical content and task lists unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Update the shared-heritage docs that still described the old combined monorepo:
D15 and the HQ console spec now note the 2026-07-16 split into two independent
repos, and the deploy runbook uses the correct repo name (sims-hq, was sims-next).
Historical decisions are preserved; only stale present-tense/naming is corrected.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Marks the split spec DELIVERED with the outcome: both repos verified, the
pre-split combined repo deleted, and the kept-@sims/kept-env decisions recorded.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds the front-page README, a first STATUS.md build-state doc (the console had
none), and a CLAUDE.md with the house rules that govern this repo. De-links four
references to Store-only planning docs (04-ROADMAP, 08-MARKET-ARCHITECTURES,
09-UX-FLOWS-AND-MENUS) that now live in the Store repo, and regenerates the
lockfile to drop the removed Store workspace entries.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This repo keeps its 60-commit HQ history and becomes independent of the Store
product. Removed the Store apps (pos, store-server, backoffice), the Store-only
packages (config, scanning, search-core, printing — printing was already a dead
dependency in hq-web), and the Store planning docs.
Shared packages are forked and trimmed to exactly what HQ uses: domain keeps
ids/money/business-day/doc-series/gstin/documents; auth keeps pin; billing-engine
keeps compute/tax. Store-only tests for the removed modules were dropped, and
auth.test.ts was trimmed to its PIN blocks. Package renamed sims-hq; root
tsconfig/vitest paths narrowed to the four kept packages.
Verified independently: npm install + typecheck clean, 177 tests pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two products share one repo: HQ (internal client book, proforma billing,
call tracking) and Store (the GST retail product shipped to ~300 clients).
Different DBs, different audiences, different cadences — 60 of 75 commits
touch HQ, 2 touch Store.
Design: two repos. HQ inherits the existing history; Store starts fresh so
its history carries zero HQ bytes. Shared packages fork into each side,
trimmed to verified usage — safe because GST rates are dated DB rows, not
code, so the forked arithmetic has no reason to diverge.
Project tracking and conversion reporting are deferred: they don't exist
yet, and mixing new features into a mechanical move hides failures.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Snapshot of the store-product workstream's in-progress work (apps/pos,
apps/backoffice, apps/store-server, packages): returns + credit notes, GST
returns exports, offline bill queue, rate-limiting, session policy, print
guard, batch tracking, and the SEC-C6/SEC-A5 auth fixes; plus docs/18 red-team
review. Committed to preserve work for the repo push. HQ console already
committed separately.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Turnkey deployment for the AWS box: documented env vars, pm2/systemd run,
HTTPS reverse-proxy, nightly SQLite->S3 backup with rotation, and the Postgres
switch note. No secrets; unblocks go-live the moment the server/creds arrive.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Automated review (CRITICAL): route registered above requireAuth, identity
fields trusted from the body. Two fix options specified; must land before
any pilot. Companion to SEC-C6.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Thomas Joise
Thomas Joise