c84f951a5c2f51279277a506754b2bd94de28249
4
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Jakob WennbergandClaude Fable 5
|
f9ef8913ae |
refactor(onboarding): extract first-year defaults + shared TIC lookup client (journey PR A) (#1141)
First of four PRs replacing the onboarding wizard with the journey flow
(dev_docs/onboarding_migration_plan.md, local). No UI change.
- Move deriveFirstYearDefaults + parseStartMonthDay out of
WelcomeOnboarding into lib/company/first-year-defaults.ts and unit-test
them (11-vs-13-months boundary, UTC month seeding, malformed input).
- Add the missing computeFiscalPeriod unit tests (calendar year, brutet
ar, first year short/extended, EF calendar-year rule, period names,
BFL 3 kap. 6-18 month window errors).
- New shared fetchCompanyLookup() client: the single client path to the
Lens-backed /lookup, typed outcomes (found / not_found / disabled /
error / aborted), never throws. Fixes the 403/404 conflation: the
dispatcher's 404 ("Extension not found") and feature-flag 503
(EXTENSION_DISABLED) now degrade silently instead of rendering as
"company not found"; only the TIC handler's own 404 does.
- Step2CompanyDetails consumes the helper; identical UX otherwise.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
||
|
Jakob WennbergandClaude Sonnet 5
|
ec27228a8e |
style: remove em/en dashes repo-wide, add CLAUDE.md rule against them (#890)
Em dashes (—) and en dashes (–) had spread across comments, docs, tests, and a few UI strings, reading as AI-generated boilerplate rather than house style. Replaced each with punctuation matching its context: colon for explanatory clauses, comma for asides, plain hyphen for numeric/legal ranges (e.g. "21-23§"), "to"/"till" for date ranges, parentheses for paired-dash asides. messages/en.json and messages/sv.json were fixed by hand together to keep sv/en in sync. Left untouched where the dash is the functional subject rather than decorative punctuation: date-range-parser.ts's separator regex, charset-repair.ts's CP1252 byte-mapping table (and its test), the SIE encoding mojibake docs, generic-csv.ts's minus-sign normalizer, the agent system-prompt files that already instruct against em dashes, and a golden iXBRL test fixture compared byte-for-byte. Also fixes two bugs surfaced along the way: an off-by-one in ApiKeysPanel's scope-label split (a leftover from an earlier partial pass), and a charset-repair test that had lost the literal en-dash it exists to verify. Regenerated the agent atom seed migration (skills:generate) since 27 SKILL.md files changed. Added a CLAUDE.md rule against em/en dashes, with an explicit carve-out for the functional-dash cases above. Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com> |
||
|
Jakob WennbergandClaude Opus 4.7
|
8fd3f112f8 |
fix: surface active TIC companies + block duplicate org numbers (#344)
* fix: surface active TIC companies + block duplicate org numbers
Three fixes from live-prod testing:
1. Enrichment filter hid the user's directorships. Now accepts both
Completed and PartiallyCompleted status from TIC (tenants without
CompanyRoles enabled still get SPAR) and the /select-company role
filter no longer requires companyStatus === 'Aktivt' — real TIC
payloads have been observed with different values, and positionEnd
alone is the authoritative "currently a director" signal. Added
PII-free diagnostic logs so the next shape-mismatch is debuggable
from Vercel logs without a round trip.
2. Manual wizard silently allowed duplicate org numbers. Added:
- findExistingCompanyByOrgNumber helper in actions.ts (service role,
bypasses RLS to see cross-tenant rows)
- Server-side guard in createCompanyFromOnboarding — returns
'org_number_exists' before the create RPC so we don't leave ghost
companies
- New /api/company/check-org-number endpoint for debounced client
checks
- Warning + disabled submit in Step2CompanyDetails
- Friendly error toasts in WelcomeOnboarding + BankIdCompanyPicker
- Mirror cleaned org_number onto companies.org_number on creation so
future duplicate checks and lookups are reliable
3. /onboarding ignored ?org_number= when the picker routed there as a
fallback. Now reads searchParams and pre-fills settings; also fixed
a latent bug where Step1's entity-type change wiped the pre-fill on
*first* selection (it should only reset on a genuine change).
Tests: duplicate-org guard (with formatted-input normalization),
check-org-number route (auth + 400 + exists true/false +
normalization).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix: address PR review feedback on duplicate-org guard
Greptile P1 findings + swedish-compliance feedback:
- findExistingCompanyByOrgNumber now throws on Supabase error instead
of silently returning null. Previously a DB outage or RLS
misconfiguration would bypass the entire duplicate guard and allow
duplicates through.
- createCompanyFromOnboarding catches the throw and returns a
user-facing error ("Kunde inte verifiera organisationsnummer"),
failing closed instead of open.
- companies.update({ org_number }) error is now checked and triggers a
rollback. Silent failure would leave the company without an
org_number, breaking all future duplicate checks for that entity.
- New normalizeOrgNumber helper validates 10- or 12-digit input,
strips the century prefix for 12-digit personnummer form, and
rejects anything else. Malformed input would have corrupted SIE4
(#ORGNR) and SRU (INFO.SRU) exports downstream.
- /select-company now uses loose `== null` for positionEnd — TIC has
been observed returning `undefined` for open-ended positions, which
strict `=== null` would silently filter out. Documented the two
downstream isCeased guards so future maintainers don't remove one
without the other.
- createCompanyFromTicRole refuses to provision when lookup.isCeased
(BFL 2 kap — bokföringsskyldighet ends at avregistrering).
BankIdCompanyPicker surfaces this client-side too.
- WelcomeOnboarding + BankIdCompanyPicker recognise new error codes:
org_number_invalid, company_ceased.
Tests: +4 cases covering malformed input rejection, fail-closed
behaviour on DB error, 12-digit personnummer normalization, and the
ceased-company refusal path. Full suite: 2306 passing.
Out of scope for this PR (follow-up):
- Partial unique index on companies(org_number) WHERE archived_at IS
NULL. Closes the race-condition window but needs a migration plus
any existing-duplicate cleanup — too risky for this hotfix.
- Rate limiting on /api/company/check-org-number. Endpoint is
auth-gated so not an immediate concern.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix: add Luhn validation and extract org-number normalization
Third round of PR review feedback (swedish-compliance):
- Add Luhn-10 check-digit validation to normalizeOrgNumber. Rejects
structurally invalid org_numbers (wrong check digit) at the boundary
instead of letting them propagate into SIE4 #ORGNR and SRU INFO.SRU,
where Skatteverket and receiving accounting systems would reject
them later anyway. Reuses the existing luhnValidate helper from
lib/bankgiro/luhn.ts (Bankgirot 10-modulen — same algorithm applies
to both Bolagsverket org numbers and Swedish personnummer).
- Extract normalizeOrgNumber into lib/company-lookup/normalize-org-number.ts
so the server action and /api/company/check-org-number use the same
rule. Previously the API route only stripped hyphens/spaces, so a
12-digit input would miss a stored 10-digit duplicate and mislead the
client debounce check ("not a duplicate" → submit → server rejects).
- /api/company/check-org-number now returns exists=false for
Luhn-invalid input rather than querying the DB. The submit-time
server action surfaces org_number_invalid, which is the right place
for the error.
Test coverage: dedicated normalize-org-number.test.ts (10 cases
covering both-lengths, Luhn, whitespace tolerance, garbage). Updated
existing tests to use Luhn-valid numbers (real Volvo 5560125790,
synthetic personnummer 8001011231). New failing-Luhn test in
actions.test.ts. New 12-digit-normalization and
luhn-invalid-returns-false tests in route.test.ts.
Full suite: 2315 passing.
Not fixed (out of scope for this hotfix):
- 10↔12 digit round-trip fragility for personnummer born 2000+. This
is a codebase-wide architectural choice (see lib/skatteverket/format.ts
which uses a two-digit-year heuristic to choose 19/20 at export).
Migrating to 12-digit storage is a separate refactor.
- Server-side re-fetch of TIC /lookup for isCeased. The trust boundary
here is user-to-their-own-onboarding, not adversarial; doubling TIC
API cost isn't proportionate.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
||
|
Jakob WennbergandClaude Opus 4.7
|
f3a3d07ed3 |
feat: one-click company setup from BankID directorships (#309)
* feat: one-click company setup from BankID directorships After BankID auth, surface Bolagsverket companies where the user is a director and provision a fully-configured gnubok company with one click instead of walking the 4-step wizard. Also exposed via CompanySwitcher's "Lägg till företag" for returning users. - New /select-company route merges gnubok memberships with TIC CompanyRoles; cards flag already-registered org numbers. - createCompanyFromTicRole server action derives entity_type, f-skatt, VAT, moms_period, and SPAR address defaults, then delegates to createCompanyFromOnboarding for consistent provisioning. - TIC /bankid/complete now requests enrichment on login too, so returning users see fresh CompanyRoles in the picker. - Middleware routes zero-membership users to /select-company when enrichment is available, /onboarding otherwise. - Inline enrichment picker removed from WelcomeOnboarding (wizard is now the manual fallback); SPAR address pre-fill preserved. - Unit tests for mapEntityType helper and createCompanyFromTicRole defaults (VAT-AB, non-VAT EF, unmappable, unauth). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix: address PR review feedback on BankID company picker Greptile P1 + swedish-compliance bot findings: - Move enrichment row cleanup out of createCompanyFromOnboarding and into createCompanyFromTicRole. The manual wizard also goes through createCompanyFromOnboarding, and was wiping the enrichment row before the returning-user "Lägg till företag" flow could use it. - Refuse to provision when TIC /lookup is missing. Silently defaulting vat_registered to false for a momsregistrerat bolag would create a company that issues invoices without moms (ML 17 kap violation). The picker now routes to the manual wizard with org_number pre-filled when the lookup fails, so the user confirms VAT/F-skatt manually. - Default accounting_method by entity type: enskild firma → cash (K1/kontantmetoden per BFNAR 2013:2), aktiebolag → accrual (K2/K3). - Document that moms_period='quarterly' is a provisional middle-tier default; Skatteverket's assigned period depends on turnover and the user can correct it in /settings/tax. - Fix the misleading "re-fetch from BankID" comment — /select-company only reads the cached enrichment row; it's refreshed only on the next BankID auth. - Extend test coverage: lookup-missing refusal, EF kontantmetoden default. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix: tighten entity-type mapping and clarify K1 threshold Second round of PR review fixes (swedish-compliance bot): - mapEntityType now uses explicit allow-lists instead of substring matches. "Enskild stiftelse" / "Enskild näringsverksamhet utan firma" no longer false-match as enskild_firma (would have provisioned with K1/kontantmetoden — ML/BFL risk). Regression guard test added. - Publikt aktiebolag explicitly included (same K2/K3 regime as private AB); Bankaktiebolag / Försäkringsaktiebolag excluded (FFFS regime). - Remove misleading claim that onboarding UI flags moms_period as provisional — no such UI exists by design (approved one-click UX). - Expand accounting_method comment to cite the 3 MSEK K1→K3 threshold (BFNAR 2013:2 vs 2017:3) so the EF→cash default is honest about its scope. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |