sax3l/accounted
3
0
Fork 0
Code Issues 1 Pull Requests 2 Actions Packages Projects Releases Wiki Activity
478 Commits 6 Branches 0 Tags
122bcbbcc12f5017a48d71dccbcbb66f877d75dd
Commit Graph

170 Commits

Author SHA1 Message Date
Joachim Hesthammer 5aa449ad3c Lock down OAuth used codes table (#641) 
* Lock down OAuth used codes table

* test(db): lock oauth_used_codes lockdown contract + reload PostgREST cache

Add a pg-real test asserting anon/authenticated are denied SELECT/INSERT on
public.oauth_used_codes while the privileged (service-role) connection can
still read it, per the project's requirement that RLS changes ship a
*.pg.test.ts. Also append NOTIFY pgrst, 'reload schema' so PostgREST picks up
the privilege change immediately (migration rule 8).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Jakob Wennberg <jakob.wennberg@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-08 09:47:33 +02:00
alfjesus 3f422b9dc8 fix(bookkeeping): restore Swedish characters in seed_chart_of_accounts (#510) 
* fix(bookkeeping): restore Swedish characters in seed_chart_of_accounts

  The seed function for new companies' chart of accounts was written with
  every å/ä/ö/Å/Ä/Ö stripped from its SQL string literals — 'Arets
  resultat' instead of 'Årets resultat', 'Avrakning socialavgifter'
  instead of 'Avräkning socialavgifter', 'Utgaende moms forsaljning' on
  the VAT accounts, and so on. Every new company has been getting 24
  starter rows with mangled Swedish names visible in the UI, reports,
  and SIE exports.

  SIE-imported accounts were unaffected because that path reads names
  from lib/bookkeeping/bas-data/. Only seeded rows looked wrong, which
  is why "SRU-mapped" imports rendered correctly while the seeded
  baseline did not.

  Diacritics are restored on the 24 affected names. The seed's
  K1-friendly short forms are kept ('Varuinköp', not BAS-canonical
  'Inköp av handelsvaror (gruppkonto)') — only the spelling is fixed.

  Two adjacent improvements in the same function definition:

  - sru_code is now populated on every seeded row from BAS reference
    data. Previously NULL on seeded rows, which broke SRU/INK2 tax
    filing for users who never import a SIE file.
  - SET search_path = public pins resolution inside the function body,
    closing Supabase linter 0011. This attribute was silently dropped
    when the seed was DROP/REPLACE'd in 20260330130000 and again in
    20260513120100 without re-declaring it.

  Existing companies are intentionally not backfilled. Only new
  companies created from this point see the fix.

  Verified with extended pg-test (9 cases, including the 2 pre-existing
  canaries) at supabase/migrations/__tests__/seed-chart-of-accounts.pg.test.ts.

  Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bookkeeping): address greptile review on PR #510

- NULL out sru_code for enskild_firma equity accounts (2010, 2013, 2018).
  BAS reference maps these to INK2 SRU 7221 ("Övrigt eget kapital"), but
  EF entities file NE-bilaga instead. Emitting `#SRU 2013 7221` in SIE
  exports would steer downstream tax software to report owner drawings
  as balance-sheet equity, which is wrong for sole traders.
- Re-add GRANT EXECUTE ON FUNCTION ... TO authenticated, restoring the
  privilege silently dropped by each DROP+CREATE since 20260330130000.
  Application paths via service-role bypass RLS and didn't break, but
  direct RPC calls from the authenticated role had been failing with
  permission errors since 2026-03-30.
- Test: lock in sru_code = NULL on the three EF equity rows.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(migrations): rename to avoid timestamp collision with PR #508

20260516120000 was already taken by assets_and_depreciation from PR #508
once that merged to main. schema_migrations uses the timestamp as primary
key, so `supabase db reset` fails with a duplicate-key error when both
files share a version.

Bump to 20260516130000 to slot between assets_and_depreciation (120000)
and depreciation_schedules_updated_at (140000).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-06-08 09:41:39 +02:00
Filip 3d3f242d11 fix(db): migrations failing on a fresh supabase instance (#645) 
* fix(db): add extension function wrappers (avoid shadowing gen_random_uuid)

Signed-off-by: filip <filip.harald@gmail.com>

* test: added test for extension function wrappers (uuid_generate_v4 and gen_random_bytes) in public schema, covering existence, delegation, correctness, properties, and idempotency.

Signed-off-by: filip <filip.harald@gmail.com>

---------

Signed-off-by: filip <filip.harald@gmail.com>
 2026-06-08 09:41:36 +02:00
Jakob Wennberg 0bc81d4c88 feat(auth): SoD acknowledge on stage+approve keys + agent:write scope for memory tools (P0-3) (#681) 
* feat(auth): SoD acknowledge on stage+approve keys + agent:write scope for memory tools

Segregation of duties on API keys is now warn + explicit acknowledgement
(not block): minting a key with any staging write scope AND
pending_operations:approve returns 409 API_KEY_SOD_CONFLICT unless the
caller re-POSTs with acknowledge_sod: true. The acknowledgement is recorded
(sod_acknowledged_at / sod_acknowledged_by) for an auditable risk acceptance
(ISO 27001:2022 A.5.3 / BFNAR 2013:2). The create UI surfaces an inline
warning and an explicit confirm dialog before submitting the ack — the
default "all scopes ticked" create routes through that path.

Also introduces the agent:write scope and maps the previously-UNMAPPED memory
tools gnubok_remember_fact / gnubok_forget_fact to it. Because unmapped tools
were callable by any key, the migration grandfathers agent:write onto every
existing non-revoked key with an explicit scope list so nothing regresses;
new keys must opt in. agent:write is deliberately excluded from the default
grants and is NOT a staging scope (no SoD conflict with approve).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(db): enforce both-or-neither on the SoD acknowledgement pair

Review finding (Greptile P2): sod_acknowledged_at/sod_acknowledged_by were
independently nullable, so a partial write could silently pass and undermine
the auditable risk acceptance (ISO 27001 A.5.3 / SOC 2 CC6.1). Adds a
paired-NULL CHECK constraint + pg-real coverage for both partial-write
directions.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* docs(auth)+feat(auth): compliance-review round — self-attestation documented, ack logged, SoD boundary assumption captured

- Migration header now states explicitly that the SoD acknowledgement is a
  SELF-attestation by deliberate design (enskild firma has no second person;
  the claude.ai approval flow needs stage+approve on one credential) — the
  control objective is informed consent + audit record, not dual control.
- The acknowledge_sod=true path now emits a structured log.warn
  (api_key.sod_acknowledged with key id/prefix, conflicting scope, scopes,
  acknowledger, company) so the acceptance lands in the logging pipeline in
  addition to the sod_acknowledged_* columns (ASVS V16.1.1).
- STAGING_SCOPES carries the documented system control (BFNAR 2013:2
  systemdokumentation) for why agent:write is not a staging scope: memory
  tools write advisory agent context and cannot stage räkenskapsinformation.

Dismissed as by-design/verified: hard-block and second-approver remediations
(user decision: warn + acknowledge); scope-update gap (the [id] route only
supports DELETE — scopes are immutable post-creation); session-auth concern
(withRouteContext is cookie+MFA only; API-key auth exists only on /api/v1
and MCP).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* chore: re-trigger CI (Supabase Preview 502 infra hiccup)

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-06 11:13:35 +02:00
Jakob Wennberg 305f469fc3 harden(db): tenant backstop — payment company-consistency triggers + write-RPC guards (P0-2) (#680) 
* harden(security): payment-row company-consistency triggers (tenant backstop)

invoice_payments and supplier_invoice_payments are the only two child tables
carrying BOTH a parent FK and their own company_id. A row whose company_id
disagrees with its parent's company_id is a tenant-isolation defect that would
surface a foreign tenant's payment in this company's AR/AP ledger. RLS scopes
by company_id but never cross-checks the parent, so nothing at the DB layer
guaranteed the invariant.

- Pre-flight DO block: fail the migration loudly (listing offending ids) if any
  existing row already violates child.company_id = parent.company_id, rather
  than arm a trigger over dirty data that can never be updated again.
- enforce_payment_company_consistency(): one INVOKER trigger function
  parameterized on TG_TABLE_NAME, wired BEFORE INSERT OR UPDATE OF
  (company_id, parent_fk) on both payment tables; raises on mismatch. Matches
  the SECURITY posture of the sibling enforcement triggers in migration 017.
- pg-real coverage in tests/pg/payment-company-consistency.pg.test.ts: matching
  pair inserts ok; cross-tenant insert + cross-tenant UPDATE raise; both the
  customer and supplier side.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* harden(security): tenant guards on six SECURITY DEFINER write RPCs (backstop)

bulk_book_transactions, match_batch_allocate, mark_entry_as_opening_balance,
reserve_voucher_range, release_voucher_range and rotate_company_inbox are all
SECURITY DEFINER and EXECUTE-able by `authenticated`, so an authenticated user
could call them via PostgREST with ANOTHER company's p_company_id. Three already
carried an auth.uid()-based membership check and rotate_company_inbox an
owner/admin gate, but the two voucher-range RPCs had NO tenant check at all.

Adds the canonical claims-based guard (mirrors
20260615120000_link_voucher_rpcs_tenant_guard.sql lines 54-69) at the top of
each body: for anon/authenticated callers, membership of p_company_id
(public.user_company_ids()) is required else RAISE 42501; service_role and
no-claims callers (migrations, pg-harness, MCP / API-key paths whose company
scoping happens in TS) bypass BY DESIGN. Each function body is otherwise copied
verbatim from its latest definition; existing GRANTs re-applied.

pg-real coverage in tests/pg/securitydefiner_write_rpc_tenant_guards.pg.test.ts:
per RPC — userA session targeting companyB raises 42501; targeting own company
passes the guard (succeeds or yields a non-42501 domain outcome, documented
inline); a no-claims bare-pool cross-tenant call bypasses the new guard,
proving the service-role / MCP paths are unaffected.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* chore(db): renumber tenant-backstop migrations to 20260619130000/130100

PR1 (agent attribution) claimed the 20260619120000 version slot in the same
batch; Supabase migration versions must be unique across the repo, so the
tenant-backstop pair moves to 130000/130100. Filename-only change plus the
matching doc-comment references in the two pg tests.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* docs(db): restore source comments dropped in copied RPC bodies

The guarded redefinitions of bulk_book_transactions and match_batch_allocate
must be byte-verbatim copies of their latest sources (modulo the inserted
tenant-guard block) so the next CREATE OR REPLACE copy keeps full provenance.
Restores the Round-2/Round-3 compliance-fix annotations that were lost in the
copy. Verified mechanically: zero residual diff vs sources after stripping the
guard block, for all six functions.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(db): drop raise-guards from bulk_book/match_batch — they break the jsonb error contract

Local full-migration replay + pg-real run surfaced that prepending the
42501 raise-guard to bulk_book_transactions and match_batch_allocate
changes their error contract for authenticated cross-tenant callers: both
already enforce membership in-function and return structured domain errors
(BULK_BOOK_UNAUTHORIZED / BATCH_UNAUTHORIZED) that routes, MCP tools, and
their existing pg tests branch on. The guard added no isolation (they were
tenant-safe) but broke that contract. The migration now guards only the
four RPCs where it is sound: mark_entry_as_opening_balance (P0001→42501,
still an exception), rotate_company_inbox (already 42501), and the two
genuinely unguarded voucher-range RPCs.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(db): compliance-review round — log hygiene, explicit INVOKER, anon revoke, UPDATE-path test

Addresses the compliance-swarm findings on this PR:
- Pre-flight dirty-data check now raises with COUNTS only; the row ids move
  to RAISE NOTICE so error pipelines do not ingest identifier dumps
  (ASVS V8.2.1 / SOC 2 CC6.1).
- enforce_payment_company_consistency() declares SECURITY INVOKER explicitly
  — the default was already INVOKER; this makes the security model
  self-documenting.
- REVOKE ... FROM PUBLIC, anon on reserve/release_voucher_range and
  rotate_company_inbox, matching the mark_entry_as_opening_balance pattern.
- Adds the missing supplier_invoice_payments UPDATE-path trigger probe
  (SOC 2 PI1.3).

Dismissed as by-design: the JWT-claim trust boundary (set_config requires
direct SQL access, which already bypasses by design — same model as
20260615120000).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(db): voucher-range compliance guards + FK-rerouting trigger probes

Review round 2 on this PR:

Swedish compliance review (both pre-existing function behaviour, hardened
while the PR owns these bodies):
- reserve/release_voucher_range now refuse closed/locked fiscal periods
  (BFL 5 kap 5§ — the sequence of a locked period is räkenskapsinformation;
  mirrors mark_entry_as_opening_balance).
- release_voucher_range asserts no verifikat exist in the released range
  before rolling last_number back (BFL 5 kap 6-7§ — never re-issue or orphan
  posted verifikationsnummer). Neither guard can fire in the legit SIE-import
  flow, which only releases numbers above its highest inserted verifikat into
  an open period — and the import caller treats a failed release as non-fatal.

Greptile P2: the UPDATE OF <parent_fk> trigger leg was never probed — added
cross-tenant FK-rerouting rejection tests for both payment tables (the
supplier company_id UPDATE probe landed in the previous commit).

Verified: full migration replay on fresh supabase/postgres + 333/333 pg-real
green on an origin/main merge (incl. merged #678).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(test): release-succeeds probe must persist — callBare rolls back

The legit-path release test asserted last_number after calling the RPC via
callBare, whose BEGIN...ROLLBACK wrapper undoes the UPDATE before the
assertion reads it (caught in CI; the local pre-push replay had validated the
branch's committed state, not the then-uncommitted test). Call the RPC
directly on the pool, like the engine pg tests do for persisting calls.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-06 10:42:41 +02:00
Jakob Wennberg 8d2ff61599 feat(bookkeeping): agent attribution into the immutable ledger layer (P0-1) (#678) 
* feat(bookkeeping): agent attribution into the immutable ledger layer

Close the three attribution gaps left after 20260618120001 (which made
commit_method record 'api_key' for MCP-relayed approvals):

- journal_entries gains nullable committed_actor_type/committed_actor_label,
  stamped by commit_journal_entry in the same draft->posted UPDATE that
  writes commit_method. The RPC gains p_actor_type/p_actor_label
  (DEFAULT NULL; prior signature dropped first to avoid PostgREST overload
  ambiguity, same technique as 20260421140000).
- write_audit_log now populates audit_log.actor_type/actor_label from
  transaction-local gnubok.actor_* GUCs set by the RPC (the established
  gnubok.allow_delete pattern). Unset GUCs COALESCE to 'user' — byte-
  identical to the column's previous effective DEFAULT for every
  pre-existing write path.
- commitPendingOperation accepts opts.actor and runs the entire executor
  inside an AsyncLocalStorage runWithActor() scope read by commitEntry(),
  so EVERY journal commit an operation makes is attributed — closing the
  documented "commitMethod only reaches create_voucher" gap. MCP approve
  passes the api_key actor + key label; web single/bulk approve pass the
  user + email.

Known limitation (documented): reverseEntry posts reversal vouchers via
direct PostgREST writes, not the commit RPC — reversals keep NULL
attribution until that path is RPC-ified (follow-up).

pg-real coverage: lib/bookkeeping/__tests__/commit-actor.pg.test.ts
(RPC param stamping, audit GUC read, transaction-locality, CHECK
rejection, immutability of the new columns, single-signature guard).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(bookkeeping): split actor-context so client bundles never see node:async_hooks

CI core-only build failed: engine.ts is reachable from client component
bundles (invoices/[id] page), and the static node:async_hooks import in
actor-context.ts cannot be chunked for the browser. Split the module:

- actor-context.ts (isomorphic): CommitActor type + a storage registry +
  getActor(). In a client bundle the registry stays empty and getActor()
  returns undefined — identical to the server-side no-scope default.
- actor-context-node.ts (server-only): owns the AsyncLocalStorage, binds it
  into the registry on import, exports runWithActor(). Imported only by the
  approval paths (commit.ts), which are never client-reachable.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-06 10:05:48 +02:00
Jakob Wennberg bc61862e76 feat(agent): telemetry + CI-gate quick wins from the "AI systems that ship" audit (#677) 
* feat(agent): telemetry completeness + durability, CI gates, commit_method provenance

Quick wins from the "Building AI systems that ship" audit:

- mcp.tool_called gains errorMessage (message_sv, truncated 500 chars) on
  all failure exits; new mcp.skill_loaded event on every gnubok_load_skill
  (all tiers) so atom usage is finally measurable
- event_log: (event_type, created_at) index; cleanup cron keeps
  mcp.*/agent.* telemetry 180 days (delivery events stay 30)
- CI: lint ratchet (npm run check:lint — 60 legacy errors baselined,
  fails only on NEW errors) and a pg-real coverage gate (migrations
  touching trigger/RPC/RLS/DEFERRABLE require a *.pg.test.ts change;
  escape hatch: -- pg-test: covered-by/skip)
- journal_entries.commit_method CHECK widened with 'api_key'/'agent';
  the MCP approve path records 'api_key' truthfully instead of
  'user_accept' (agent_first_vision §8 P0-1). 'agent' is reserved — ALL
  MCP traffic (incl. claude.ai OAuth, whose access_token is a minted
  API key) authenticates as api_key today

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(import): derive opening balances from prior-year #UB when SIE lacks #IB (#675)

SIE files exported without #IB 0 rows (only #UB -1) previously imported
with zero opening balances. getEffectiveOpeningBalances() now derives IB
from prior-year UB for balance-sheet accounts when explicit #IB is
absent, surfaces the derivation as an info issue in the import preview,
and excludes share-capital vouchers from opening-balance detection.
Detection regexes are shared between parser and importer so the two
checks cannot drift. 507 lib/import tests pass.

(Authored in a parallel session in this checkout; included per request.)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(review): address PR #677 bot findings — RoPA entry, execFileSync, gate scope note

Triage of the compliance-swarm + Greptile findings:

Applied:
- .compliance/ropa.yaml: new mcp.telemetry processing activity declaring
  the 180-day mcp.*/agent.* retention, lawful basis, data categories, and
  the no-args/no-results minimisation (ISO A.8.10, GDPR Art.5(1)(c) —
  the retention split is now formally documented, referenced from the cron)
- check-pg-test-coverage.mjs: execFileSync with argv array — no shell, so
  a hostile base-ref can't inject (ASVS V13.2.1); verified an injection
  attempt exits 2 without executing
- check-pg-test-coverage.mjs: documented the PR-level (not per-migration)
  scope of the gate so reviewers know to check coverage per migration when
  a PR carries several risky migrations (Greptile P2)

Acknowledged, no change:
- errorMessage PII risk: messages are domain-mapped strings; event_log
  already persists far richer delivery payloads under the same RLS; now
  declared in ropa.yaml
- cron error envelope: errorResponse maps to the canonical safe envelope
  and the endpoint is CRON_SECRET-gated
- two-pass delete "partial state": TTL deletes are idempotent — the next
  daily run sweeps whatever a failed pass left behind
- skill_loaded actorLabel/sessionId: mirrors the pre-existing
  mcp.tool_called payload; sessionId is the join key the analytics exist for

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-05 15:47:13 +02:00
Jakob Wennberg f59da07fc0 feat(worklist): unified pending-work counts + vampire-transaction fix (#673) 
* feat(worklist): unified pending-work counts in lib/worklist

One source of truth for every "Att göra" count, shared by the sidebar
badges, the home page (PR 3), and eventually the MCP list tools. Each
category documents its pending/done predicate in types.ts; counts are
cheap head-only queries that soft-fail to 0 so a broken badge can never
take down a layout. listSuggestedMatches() returns confirmable
transaction↔invoice match rows for the upcoming one-click confirm UI.

GET /api/worklist/counts exposes the aggregate for client refetch.

Partial index on transactions (company_id WHERE unbooked, not ignored)
— the badge predicate runs on every dashboard render — and a
company-scoped index on invoice_inbox_items, which was only indexed by
user_id from before the multi-tenant refactor.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(transactions): unify unbooked predicate, add Ignorera action

The "vampire transactions" fix. Three surfaces counted "unbooked"
differently — the sidebar badge included ignored rows (is_business IS
NULL only), the transactions page count and inbox list disagreed with
each other — so old transactions kept haunting the inbox with inflated
badges after every sync. All three now share lib/worklist's canonical
predicate: is_business IS NULL AND is_ignored = false.

The transactions page also gains "Ignorera transaktionen…" in the
booking dialog (mirroring BankReconciliationView's flow: confirm →
ignore → Ångra toast), so non-business strays can finally be cleared
from the worklist without fabricating a verifikation. Recovery remains
on Rapporter → Bankavstämning under "Ignorerade transaktioner".

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(worklist): address PR #673 review — exact missing-underlag count, tenant-tagged logs

Review findings:
- countVerifikatMissingDocument subtracted set SIZES, which both let
  documents on non-document-requiring entries (e.g. VAT settlements)
  shrink the count and silently truncated at the PostgREST row cap on
  large document histories. Now an exact per-entry set difference over
  fetchAllRows-paginated id-only reads. (greptile, swedish-compliance)
- logAndZero gains companyId as a structured log field so repeated
  count failures can be correlated to a tenant. (compliance-swarm V16)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(worklist): address PR #673 review round 2 — chunked .in(), BFL-aware ignore copy

- countInboxDocuments: dedupe + chunk the document-id .in() filter at 150
  ids per request — PostgREST serialises it into the GET query string, so
  a large inbox could exceed proxy URL limits (HTTP 414) and silently
  zero the badge via the error branch. (greptile)
- Ignorera confirm dialog now states the BFL boundary: only for
  non-affärshändelser (duplicates, own-account transfers) — real
  purchases and payments must be booked. Persisting a structured ignore
  reason is noted as follow-up. (swedish-compliance)
- New tests: chunked counting sums across batches; a mid-stream
  pagination error in countVerifikatMissingDocument throws (fetchAllRows
  propagates page errors) and soft-fails to a logged 0 — never a count
  computed from a silently truncated set. (swedish-compliance)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-05 10:11:37 +02:00
Mattsson 4a54467599 Bug/transaction date corruption (#668) 
* fix(transaction): enforce valid date range for transactions and add database constraint

* fix(transaction): implement server-side validation for transaction dates and enhance error handling
 2026-06-04 16:16:27 +02:00
Jakob Wennberg f538401988 Invoice correctness bundle: voucher-link race, agent send guards, payment-reversal restore (audit C2/C17, F-2026080) (#666) 
* fix(invoices): atomic link_invoice_to_voucher RPC — close the customer voucher-link race (audit C2)

linkInvoiceToVoucher() did UPDATE-then-INSERT with a manual rollback restoring a STALE pre-link snapshot: under concurrent linking on the same invoice, A's failed insert could overwrite B's successful link while B's payment row remained — corrupting paid_amount/AR. Mirrors the supplier-side link_supplier_invoice_to_voucher fix (PR #602).

- New SECURITY DEFINER RPC locks the invoice FOR UPDATE, re-validates (status, posted voucher, 151x AR credit, currency, overshoot, already-linked) and applies UPDATE + INSERT in one PG transaction. Inherits the supplier RPC's remaining-amount fix (trust stored remaining_amount even at 0 — the TS '> 0' guard let rounding drift slip past FULLY_PAID). Hardened per audit A5: REVOKE from PUBLIC/anon, GRANT to authenticated + service_role.
- linkInvoiceToVoucher() now delegates to the RPC — same signature, same LINK_VOUCHER_* codes, so all callers (route, pending-op executor, MCP) are unchanged. Keeps the invoice.paid event (now emitted with the post-link row, mirroring the supplier wrapper) and the best-effort bank auto-reconcile.
- pg-real tests: full/partial link, overshoot leaves the invoice untouched, ALREADY_LINKED, and the race regression (two concurrent full links -> exactly one wins, paid_amount never exceeds total, exactly one payment row). Verified locally against supabase/postgres:15.8.1.060 with all 334 migrations replayed: 10/10 pass. Two unrelated pg tests fail locally with AND without this change (pre-existing env sensitivity; green in CI).
- Unit tests re-mocked to the RPC-wrapper contract.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(invoices): agent send path — block cancelled invoices + preflight PDF render (audit C17)

commitSendInvoice (the agent/MCP path) was missing two guards the send route has:

- No cancelled guard: a cancelled invoice passed the already-sent check, got re-rendered and EMAILED (a 'MAKULERAD' PDF delivered as if live), and the unguarded status flip silently re-activated it to 'sent'. Now rejected with the registry's INVOICE_SEND_CANCELLED message (400), mirroring the route.
- No preflight render: the executor assigned the F-series number BEFORE rendering, so a render failure left a numbered-but-never-issued invoice (an F-series gap if the draft is abandoned). Now mirrors the route: on fresh allocation, render with an 'F-PREVIEW' placeholder first and reject with INVOICE_SEND_PDF_RENDER_FAILED before any number is consumed; retries with an existing number skip the preflight.

Items/credit-note lookup moved above the preflight (it needs them); the real render and everything downstream are unchanged.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(bookkeeping): payment reversal restores invoice state and releases bank line (F-2026080)

Reversing a payment voucher left the customer invoice deadlocked: status
stayed 'paid' while remaining_amount stayed stale (= total), and the bank
transaction kept pointing at the reversed JE so the line could neither be
re-matched nor deleted.

- Customer branch now recomputes remaining_amount from total (the supplier
  branch already did) and clamps paid_amount at 0.
- Both branches delete the payment row(s) tied to the reversed voucher so a
  re-match doesn't double-count or trip the unique indexes.
- New releaseLinkedTransactions() detaches bank transactions from the
  reversed JE (by journal_entry_id and by captured payment transaction ids),
  clearing the link/categorization columns so the line returns to the inbox.

Covers every standalone storno path (reverse route, MCP reverse tool,
delete-last-voucher); the match-invoice route already handled its own case.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(transactions): match-invoice preview double-subtracted VAT on per-item path (F-2026080)

InvoiceItem.line_total is the NET line amount (it sums to invoice.subtotal,
each line's vat_amount = line_total * rate), but the preview's per-item rate
aggregation computed sub = line_total - vat_amount, double-subtracting VAT
and producing an unbalanced previewed verifikat (revenue credit too low
against the 1930 debit). The commit path (generatePerRateLines) was already
correct; only the preview disagreed.

Regression test mirrors the F-2026080 invoice: multi-item 25% SEK cash entry
must balance, with 3001 = subtotal and 2611 = vat_amount.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(bookkeeping): address PR #666 review — supplier cash reversal, RPC tenant guard, CI fixes

Review feedback fixes:

- Supplier cash-payment reversal (Greptile): the supplier branch required a
  payment row before restoring status/amounts, so reversing a
  supplier_invoice_cash_payment (which books no payment row) left the invoice
  deadlocked at paid/remaining=0 — the same bug the customer branch fixed.
  Mirror the customer fallback (revert full paid_amount when no row exists).

- Payment-row lookups now filter by invoice id + company_id: a batch voucher
  (match_batch_allocate) carries one payment row per invoice under the same
  journal_entry_id, so the unfiltered .single() errored out and silently
  yielded null.

- Tenant guard on the voucher-link write RPCs (compliance V8.2.1, audit A5):
  link_invoice_to_voucher and link_supplier_invoice_to_voucher are SECURITY
  DEFINER + authenticated-executable, so any signed-in user could mutate
  another tenant's invoices via PostgREST. New migration applies the PR #625
  claims-based membership guard to both, caps p_notes at the Zod layer's
  2000 chars, and gives the supplier RPC the explicit REVOKE/GRANT it never
  had (was default PUBLIC execute). Covered by a new pg-real test.

- releaseLinkedTransactions now logs Supabase errors (compliance V16.1) —
  a failed release leaves a bank line stuck on a reversed JE and must be
  observable.

CI fixes:

- naive-ore-round ratchet (core-only): payment-sync.ts converted to
  roundOre() from @/lib/money (-4 occurrences vs baseline).
- match-batch-allocate.pg.test.ts flake (pg-real): Date.now()+random arrival
  numbers collided in CI; now time-component + monotonic counter.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(bookkeeping): address PR #666 review round 2 — payment attribution, batch-scoped deletes, send guard

- RPC payment attribution (GDPR Art.32): user-session callers can no longer
  attribute invoice_payments / supplier_invoice_payments rows to an arbitrary
  user via p_user_id — the JWT sub is authoritative when role is
  anon/authenticated. service_role / direct callers keep p_user_id verbatim
  (their scoping happens in TS). pg-real test asserts the spoofed id is
  ignored.

- Payment-row deletes scoped to the source invoice (SOC 2 CC6.3): a batch
  voucher carries sibling payment rows for other invoices whose status this
  sync doesn't restore; deleting them desynced paid_amount from the rows.

- releaseLinkedTransactions success audit log: transactions has no
  write_audit_log trigger, so clearing the link/categorization columns now
  logs the affected transaction ids for incident reconstruction.

- commitSendInvoice guard extended with partially_paid/credited (ASVS V2.3):
  both imply the invoice was already issued; the status flip would have
  regressed them to 'sent'.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-04 15:32:39 +02:00
Mattsson 3e42fc6f32 Feat/voucher docs (#664) 
* feat: implement inbox document picker and linking functionality

* feat: implement self-billing invoice functionality

- Added support for registering self-billed invoices received from customers.
- Updated the invoice schema to include fields for self-billing metadata such as `is_self_billed`, `external_invoice_number`, `self_billing_agreement_ref`, and `received_date`.
- Created API route for handling self-billed invoice submissions, including validation and error handling.
- Implemented database migrations to add necessary columns and constraints for self-billing invoices.
- Developed tests to ensure correct behavior of self-billing invoice creation and validation rules.
- Updated Swedish localization files to include new terms related to self-billing.

* feat: enforce SIE import requirement for non-Fortnox providers in migration process

* feat: streamline invoice processing and enhance error logging across APIs
 2026-06-04 13:14:57 +02:00
Jakob Wennberg c74b19df1b Accounted rebrand + swarm-skill cleanup + bank-reconciliation fixes (#643) 
* feat(reconciliation): close the bank-feed loop on voucher links and re-tag mis-typed opening balances

Two related fixes to bank reconciliation correctness:

1. Auto-reconcile on voucher link. Linking an invoice or supplier invoice to
   an existing voucher previously advanced only the invoice — the bank
   transaction that paid it kept sitting in the Transactions inbox with a null
   journal_entry_id. linkInvoiceToVoucher / linkSupplierInvoiceToVoucher now
   call autoReconcileTransactionForLinkedVoucher (lib/reconciliation), which
   links the bank transaction to the same verifikat when exactly one unbooked
   line matches it. Best-effort and post-commit: a failure here never fails the
   link. The result surfaces reconciledTransactionId; the inbox row leaves the
   list and the UI shows link_success_tx_reconciled.

2. Re-tag mis-typed opening balances. getReconciliationStatus and the GL-line
   matching RPCs identify a cash account's ingående balans solely by
   journal_entries.source_type='opening_balance'. Companies migrated from other
   systems often booked the bank IB as an ordinary voucher (source_type
   'import' or 'manual'), so it was never excluded and surfaced as a phantom
   reconciliation difference equal to the opening balance. Adds:
   - migration mark_entry_as_opening_balance: a GUC-gated carve-out in the
     immutability trigger plus a SECURITY DEFINER RPC that validates the entry
     (balance-sheet lines only, dated on a fiscal-period boundary), flips the
     source_type, and writes an audit row — no blanket data sweep.
   - POST /api/reconciliation/bank/mark-opening-balance + MarkOpeningBalanceSchema.
   - BankReconciliationView action to trigger it from the IB diff.

The gnubok_create_voucher executor now accepts a typed is_opening_balance flag
and derives source_type='opening_balance' only after validating class 1/2 lines
on the period start, so new IBs land correctly typed.

Covered by lib/reconciliation auto-reconcile tests, voucher-executors tests,
and a mark-entry-as-opening-balance pg-real test.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* chore: rebrand gnubok → Accounted and prune swarm agent skills

Product rebrand and skills housekeeping. No runtime behaviour change.

Rebrand: replace user-visible "gnubok" with "Accounted" across docs, READMEs,
in-code comments, doc-site content, MCP skill/resource prose, and the
gnubok-mcp package description. The MCP resource URI scheme is moved gnubok://
→ Accounted:// consistently across resource registrations, the event-type
comment, and the resource/skill tests. Deliberately preserved as stable
identifiers (NOT rebranded): the gnubok-company-id cookie, gnubok_sk_ / gnubok_inv_
token prefixes, the gnubok-mcp npm bridge name, and the AGI <gem:Programnamn>
value (kept 'gnubok' per its source comment — it is the software identifier sent
to Skatteverket and must not churn across visual rebrands).

Skills: remove the 27 swarm-* agent SKILL.md atoms (no longer used; already
absent from the agent_atom_registry in prod), refresh the remaining skill docs,
add the .claude/rules/ path-scoped rule set, and regenerate the
seed_agent_atom_bodies migration + .skill-body-manifest.json via
`npm run skills:generate` so the DB-backed skill bodies match the trimmed set.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-03 10:52:01 +02:00
Mattsson f6ee0c2a82 Bug/customer invoice bug (#628) 
* fix(supplier-invoices): self-assess reverse-charge VAT + link payments to vouchers

Reverse-charge supplier invoices now carry a per-item reverse_charge_rate (0.06/0.12/0.25). Under omvänd skattskyldighet the supplier charges 0% VAT, so the line vat_rate stays 0 and the buyer self-assesses fiktiv moms at the statutory rate. Centralizes rate resolution (resolveReverseChargeRate) and the ruta 20-24 basis-account guard (isReverseChargeBasisAccount) in vat-entries so the booking engine and review-dialog preview can no longer drift.

Adds the link_supplier_invoice_voucher pending operation: mark a leverantorsfaktura paid by linking an existing posted verifikat that debits 2440, with no new journal entry. Exposes find-candidates/link MCP tools and the bulk-reconcile helper, scoped under suppliers:read/write.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(vat): report yearly VAT over the rakenskapsar, not the calendar year

Annual VAT (helarsmoms) is filed per beskattningsar/rakenskapsar (SFL 26 kap), which can be extended or shortened up to 18 months. The previous Jan-Dec calendar span silently dropped part of an extended first year. calculateVatDeclaration now accepts a fiscalPeriodId and resolves the period's actual bounds for yearly; monthly/quarterly stay calendar. The reports UI passes the selected fiscal period, defaults the periodicity from the company's moms_period setting, and carries the period into the ruta drill-down. full-archive export threads the period id through too.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(migration): resolve supplier invoice status from payment amounts

The provider's lifecycle status and its payment status are computed independently upstream and can contradict each other (e.g. a Fortnox invoice marked booked but fully paid). Both the arcim entity-mapper and the Fortnox mapper now let payment state win: fully paid -> paid, partial -> partially_paid, otherwise the mapped lifecycle status, with credit notes forced terminal. Balance is compared numerically (never strict === 0) so float drift or a residual ore resolves cleanly, and an absent Balance is treated as unpaid.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(enable-banking): only ingest booked transactions to stop re-import drift

Pending entries are skipped during sync: a pending row is unstable across syncs (a later 'synka nu' returns it still pending or finally booked, often with a different effective date). Because both the dedup external_id and the content-dedup key are date-derived, that drift minted a new id and re-imported a transaction that already existed - observed in production as the same amount+description landing twice with different dates. Gating the import set on a stable booking_date removes the drift at the source and leaves booked rows' ids byte-identical.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* chore(gitignore): ignore local SIE test fixtures

tests/fixtures/sie/ may contain real or scrubbed company data and must never be committed.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(invoice): handle errors during registration journal entry creation and ensure invoice rollback
feat(tests): add test for reverse charge rate handling on supplier invoice line items
feat(fortnox): ensure paid status reflects zero balance for fully paid invoices
chore(migrations): add reverse_charge_rate to supplier_invoice_items and backfill link_supplier_invoice_voucher

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-02 13:25:48 +02:00
Jakob Wennberg 358c25094d fix(security): tenant guard on the GL-line read RPCs (#625) 
* fix(security): tenant guard on the GL-line read RPCs (PR #624 follow-up)

get_unlinked_gl_lines and get_account_gl_lines_for_matching are SECURITY DEFINER
and EXECUTE-able by anon/authenticated, so any authenticated (or anonymous)
caller could invoke them directly over /rest/v1/rpc with another company's id and
read its general-ledger lines — a cross-tenant read that bypasses the API routes'
requireCompanyId() guard. Confirmed against the DB: anon and authenticated both
hold EXECUTE, and SECURITY DEFINER sidesteps RLS.

Add an in-function guard constraining anon/authenticated callers to their own
companies (the same boundary user_company_ids()/RLS enforces). Trusted callers
are untouched — service_role (the enable-banking reconciliation cron) and direct
/ superuser access (migrations, the pg-real harness) are not anon/authenticated,
so the predicate is a no-op and behaviour is unchanged. A foreign company id now
yields zero rows, not data.

Scope: hardens the two READ RPCs that expose ledger data. The remaining
company-scoped SECURITY DEFINER RPCs are writes / sequence generators with their
own internal authorization; a broader audit of that set is tracked separately.

pg-real coverage: a company-B member probing company A gets zero rows from both
RPCs, while a company-A member and direct/superuser access still see the data.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(security): revoke EXECUTE from PUBLIC/anon on the GL-line read RPCs

Defense-in-depth follow-up to the tenant guard. The guard already returns zero
rows to an anon/authenticated caller probing another company; this additionally
strips the EXECUTE privilege so an unauthenticated (anon) caller cannot invoke
the financial-ledger RPCs at all. Supabase grants EXECUTE to PUBLIC as well as to
anon, and anon is a member of PUBLIC — revoking only anon is insufficient, so
revoke both, then keep the two legitimate callers: authenticated (the API routes
call via the user's session; the in-function guard scopes them to their own
companies) and service_role (the enable-banking reconciliation cron).

Verified on the DB: anon EXECUTE = false, authenticated/service_role = true.
Adds an anon-role pg test asserting the call is rejected at the privilege layer.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(security): read JWT role from claims object in the GL-line RPC guard

The 20260611120000 guard used auth.role() to detect the caller's role, but
auth.role() reads the individual request.jwt.claim.role GUC first and only some
installs fall back to the claims object. PostgREST sets the claims OBJECT (the
individual claim.* GUCs are deprecated), and the pg-real harness sets
request.jwt.claims (+ claim.sub for auth.uid()) but NOT claim.role — so on an
auth.role() without the object fallback it returns NULL and the guard's
NOT IN ('anon','authenticated') branch was TRUE, skipping the membership check.
A pg-real test caught it: an authenticated non-member could still read another
company's GL lines (the guard failed open in that environment).

Read the role straight from request.jwt.claims (exactly what auth.role() itself
falls back to), so the guard enforces in every environment regardless of which
JWT-claim GUCs are populated. Verified on the DB: an authenticated non-member
evaluates both guard branches false → row excluded.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-02 10:24:18 +02:00
Jakob Wennberg 094bd85e81 fix(reconciliation): secondary-account scoping, dialog clipping, and N:1 matching (#624) 
* fix(reconciliation): secondary-account scoping, dialog clipping, and N:1 matching

Three follow-ups to per-account bank reconciliation (PR #623):

- Secondary same-currency accounts (e.g. a 1931 savings account) double-counted
  the company's unassigned (NULL cash_account_id) transactions, inflating their
  bank total and showing a large bogus difference while 1930 still reconciled.
  Only the primary cash account now claims NULL rows; every other account scopes
  strictly to its own id. `includeUnassigned` is threaded through all
  status/run/list call sites from cash_accounts.is_primary.

- The "Matcha mot befintlig verifikation" picker's dropdown was absolutely
  positioned inside the dialog's overflow-y-auto container and got clipped. Add
  an `inline` mode that renders the candidate list in normal flow; the dialog
  uses it, the reconciliation view keeps the compact overlay.

- N:1 matching: several bank transactions can now settle one verifikat (a salary
  run paid in multiple transfers, an invoice paid in instalments). New
  get_account_gl_lines_for_matching RPC surfaces already-matched vouchers with a
  linked_transaction_count behind a "Visa även matchade verifikationer" toggle;
  manualLink's 1:1 guard is relaxed (the aggregate difference still catches
  mis-links).

Tests: extended bank-reconciliation unit tests (strict scope + N:1), rewrote the
cash_account_id isolation pg test to prove NULL rows land on the primary account
only, and added a pg test for the new RPC.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(reconciliation): address PR review — accurate "att matcha mot" count

- BankReconciliationView: the "N verifikationer att matcha mot" hint counted
  glLines (which includes already-matched vouchers when "Visa matchade" is on),
  overcounting the vouchers that still need a transaction. Use unmatchedGlLines
  so the label is correct regardless of the toggle (matches the table below).
- MatchVerifikationPicker: document that `open` is overlay-only; the setOpen()
  writes are intentional no-ops in inline mode.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-02 09:37:30 +02:00
Jakob Wennberg b5c3c3ec04 fix(reconciliation): surface bank tx + match to existing verifikat, delete UX, DPA links (#623) 
* fix(reconciliation): surface bank transactions via robust cash_account_id scoping

The per-account reconciliation scoping silently returned zero transactions for
companies whose rows were NULL or mis-assigned mid-backfill (e.g. Arcim: 138
transactions, 101 unbooked, yet Bankavstämning showed "0 kr" while the 1930 GL
movement and a large difference still displayed). Two causes, both fixed:

- scopeTransactionsToAccount used a fragile nested or(...,and(is.null,...))
  PostgREST filter. Replace it with a flat, reliable
  `currency = X AND (cash_account_id = id OR cash_account_id IS NULL)` and share
  the one implementation with /api/transactions so the status card and the lists
  can never drift.
- The original best-effort backfill only touched NULL rows and an earlier
  revision mis-assigned cash_account_id (the since-fixed min(uuid) bug), which
  migrations cannot self-correct. Add an idempotent repair migration that
  re-seeds the default 1930 account and re-derives cash_account_id (correcting
  non-NULL mis-assignments) for booked rows and single-account companies.

Also localise manualLink's user-facing errors to Swedish. Adds unit coverage for
the new filter shape and pg-real coverage for the repair (incl. the Arcim
single-account reproduction).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(transactions): match a bank transaction to an existing verifikat

Adds a "Matcha mot befintlig verifikation" action to the Transactions inbox so a
bank line that is already booked elsewhere (a salary run, a Fortnox/manual
voucher, an invoice paid from the invoice page) can be linked to that existing
verifikat with no new bokföring — the capability previously lived only in
Reports → Bankavstämning.

- Extract the searchable MatchVerifikationPicker into a shared client component.
- New MatchVoucherDialog: resolves the tx's cash account, fetches candidates
  ranked server-side by reconciliation confidence, links via
  /api/reconciliation/bank/link (so the link is undoable in Bankavstämning).
- unmatched-entries route gains an optional transaction_id that ranks candidates
  (ranking stays server-side; the recon lib is not client-safe).
- Inbox row's overflow (⋯) menu gains the new action.

Also fixes the Bankavstämning view: editing the date no longer auto-reloads
(applies on Filtrera / account change only) and Datum till defaults to today.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(transactions): actionable Swedish errors when a bank tx cannot be deleted

The delete route returned a hardcoded English 409 that getErrorMessage mapped to
the misleading generic "En konflikt uppstod. Ladda om sidan...". Return
structured bilingual envelopes instead:

- TRANSACTION_DELETE_BOOKED (409) for a booked/linked row — steers the user to
  unlink in Bankavstämning or storna the voucher.
- TRANSACTION_DELETE_HAS_AUDIT_TRAIL (409) for the real, common case where an
  unbooked row carries payment_match_log rows: the cascade hits the
  audit-immutability trigger (P0001), previously surfaced as a bare 500. Steers
  the user to match-to-voucher or ignore instead.

Updates the DELETE test suite and adds the P0001 case.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(settings): surface DPA and privacy policy links

The /dpa page (personuppgiftsbiträdesavtal, GDPR Art. 28) was complete but
linked from nowhere. Add a "Sekretess och avtal" card on Inställningar → Konto
linking to /privacy and /dpa, and a reciprocal link to the DPA from the privacy
policy's sub-processor section. (The DPA already links back to /privacy.)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(review): address PR #623 feedback

- unmatched-entries: when transaction_id is supplied but resolves to no row in
  the caller's company, return an empty candidate list instead of silently
  falling back to the full unranked list (Compliance Swarm V8.2.1, high).
- MatchVoucherDialog: preserve a manually-picked voucher when the candidate
  list reloads (e.g. "Visa alla datum") instead of discarding it (Greptile P2).
- DELETE /api/transactions/[id]: return the 404 as the structured
  { error: { code, message, message_en } } envelope like the handler's other
  errors, for a uniform contract (Greptile P2). Test updated.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(bookkeeping): allow editing notes on a committed journal entry

Saving a note on a posted verifikation failed with "Committed entries are
immutable": enforce_journal_entry_immutability() had no posted→posted path, so a
notes-only UPDATE fell through to the final RAISE. `notes` is internal annotation
metadata (not verifikation content under BFL 5 kap. / BFNAR 2013:2), so add a
narrow carve-out that permits a notes-only change on a committed entry — verified
with a whole-row to_jsonb() diff so any other field change still raises, and only
when status is unchanged. Period-lock enforcement is unaffected.

CREATE OR REPLACE in a new migration (same pattern as
20260428160000_fix_journal_entry_immutability_delete_bypass); the migration-017
protections are extended, never weakened. Covered by a pg-real test asserting a
notes edit succeeds while amount/description/account edits still fail.

(Already applied to production; committing the file + test for repo consistency.)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(sandbox): make pre-staged pending operations executor-complete

The seeded pending_operations existed only as display previews — approving them
failed because commit executors in lib/pending-operations/commit.ts validate
required fields on "Godkänn". Seed a backing invoice_inbox_items row and fill the
supplier-invoice and categorize params with every field the executors require
(inbox_item_id, full items array; real uncategorized transaction_id + category),
so the sandbox approval queue is actually approvable end to end.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-01 21:08:06 +02:00
Jakob Wennberg 953980c875 Per-account bank reconciliation + overdue/inbox/privacy fixes (#619) 
* feat(reconciliation): scope bank reconciliation per cash account via transactions.cash_account_id

A company with two same-currency cash accounts (e.g. checking 1930 + a
savings account) saw every SEK transaction on every account, and the
status card summed across both — reconciliation filtered transactions by
CURRENCY while filtering GL lines by ACCOUNT (issue #604).

Bind each bank transaction to the cash_accounts row it settled on:

- New nullable transactions.cash_account_id FK (ON DELETE SET NULL —
  a bank transaction is räkenskapsinformation, BFL 7 kap, and must
  survive cash-account deletion) + a best-effort 4-pass backfill.
- All reconciliation/transaction queries scope to the selected account
  with a NULL->currency fallback, so legacy/un-backfilled rows never
  disappear mid-backfill.
- ingestTransactions stamps cash_account_id from the batch's
  settlementAccount; categorize + manualLink resolve and use it.
- Bank leg now books to the transaction's actual settlement account via
  applySettlementAccount (no-op for 1930), so interest/fees on a
  savings/EUR account reconcile instead of mis-booking to 1930.
- manualLink cross-checks the transaction's account and requires a
  voucher line on the selected account (no silent cross-account links).
- BankReconciliationView: quick-book menu for any settlement account,
  in-flight request abort on account/date switch, 500-row truncation
  notice, per-account state reset.
- pg-real coverage for the FK, all backfill passes, account-scoped
  query isolation, and cross-company isolation.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(supplier-invoices): stop marking paid invoices and credit notes as overdue

update_overdue_supplier_invoices() (the daily pg_cron job) flipped every
past-due 'registered'/'approved' row to 'overdue' without looking at the
outstanding balance. Credit notes — created 'registered', remaining 0,
due today — got flipped the next day, surfacing as "Förfallen" with
"kvar att betala 0 kr"; so did any fully-paid invoice left in
'registered'/'approved'.

Guard the cron on remaining_amount > 0.005 (the "fully paid" threshold
used by the payment/match paths) and is_credit_note = false, and backfill
the rows already mis-flagged (credit notes -> 'registered', paid ->
'paid' with paid_at stamped only when missing). pg-real coverage for the
guarded function and the one-off backfill.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(invoice-inbox): refresh dokumentinkorg on realtime row changes

The InvoiceInboxWorkspace only refetched on mount and on explicit
in-component actions. When an inbox item was resolved out of band — the
in-app agent sheet committing a staged create_supplier_invoice_from_inbox
/ book-direct op, the /pending page approving one, or another tab booking
it — none of those paths called fetchItems(), so the booked underlag
stayed in "Att göra" until a manual reload (issue #600).

Add invoice_inbox_items to the supabase_realtime publication (mirrors the
/pending fix in 20260520120100) and subscribe in the workspace, refetching
the whole list on any change so derived status/counts/ordering stay
authoritative. RLS scopes the channel to the user's company. fetchItems
now preserves optimistic upload placeholders so a refetch firing
mid-upload can't drop an in-flight row.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* docs(privacy): disclose EU AI inference via Amazon Bedrock (eu-north-1)

Update the privacy policy and DPA to state that AI inference, when AI
features are enabled, runs inside the EU via Amazon Bedrock (eu-north-1,
Stockholm) using Anthropic's Claude models — no transfer to a third
country, prompts not retained after the call or used for model training.
Add AWS as a subprocessor row and refresh the "last updated" dates.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(migrations): rename invoice_inbox_realtime to avoid version collision

main's #617 shipped 20260605120000_transactions_original_description.sql —
the same version this branch used for the inbox-realtime publication. The
Supabase migration tracker keys on the numeric version, not the filename, so
the preview branch failed with a duplicate-key error on
supabase_migrations.schema_migrations (version 20260605120000 already
exists). Rename to the unique version 20260605120500; the body
(ALTER PUBLICATION) is order-independent.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(reconciliation): align run guard with status; harden filter interpolation

Addresses PR review (greptile + compliance swarm):

- The v1 and core bank/run routes rejected an unknown account uniformly,
  including the default '1930', while the status routes were lenient for
  '1930'. A company reconciling its primary SEK account without a
  cash_accounts row got 200 from status but 400 from run. Make run match
  status: '1930' falls back to currency-only scoping (cashAccountId
  undefined); non-default unknown accounts are still rejected. Adds a test.
- /api/transactions accepts a user-supplied `currency` query param that was
  interpolated raw into a PostgREST .or() filter. Reject anything that isn't
  a 3-letter ISO code — RLS already scopes to the company, but an
  unsanitized value could otherwise malform/widen the filter. Assert
  currency/cashAccountId shape in scopeTransactionsToAccount as well.
- categorize: log (instead of silently swallowing) a cash_accounts
  settlement-account lookup error, so a fall-back-to-1930 mis-booking is
  observable in the audit log.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(migrations): correct backfill UPDATE..FROM join; idempotent realtime publication

Two SQL errors that only surface on real Postgres (CI pg-real + Supabase
preview) — the unit suite mocks Supabase, so neither was caught locally.

- Backfill pass (a): `UPDATE transactions t ... FROM journal_entry_lines jel
  JOIN cash_accounts ca ON ca.company_id = t.company_id` referenced the UPDATE
  target `t` inside the FROM join's ON clause, which Postgres rejects ("invalid
  reference to FROM-clause entry for table t"). Move the company match to WHERE;
  the JOIN now relates jel<->ca only. Semantics unchanged.
- invoice_inbox_realtime: `ALTER PUBLICATION ... ADD TABLE` is not idempotent
  (SQLSTATE 42710 if the table is already a member). The earlier
  version-collision push partially applied it on the Supabase preview branch, so
  the re-apply errored. Guard with a pg_publication_tables existence check.

Both statements validated against a real Postgres: the single-line tx binds, the
two-bank-line transfer stays NULL, and the publication add runs twice cleanly.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(migrations): backfill pass (c) uses array_agg, not min(uuid)

Postgres has no min() aggregate for uuid, so pass (c)'s min(id) raised
"function min(uuid) does not exist" on apply (CI pg-real + Supabase). The
HAVING count(*) = 1 already guarantees one row per group, so (array_agg(id))[1]
returns that single id.

Validated the full backfill (all four passes) and the overdue migration against
a real Postgres: every pass binds / falls through as intended, and the overdue
guard + backfill produce the right statuses.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* docs(compliance): add RoPA entry for Amazon Bedrock AI inference (GDPR Art.30)

The privacy policy now discloses AI inference (transaction categorization +
document/receipt OCR) via Amazon Bedrock as a processing activity, but
.compliance/ropa.yaml had no matching Art.30 record. Add it: opt-in consent
basis, EU-region (eu-north-1) inference with no third-country transfer, prompts
not retained or used for model training. Mirrors the privacy-page disclosure
shipped in this PR.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-01 18:26:13 +02:00
Mattsson c6c86cded4 Mcp/template data feedback (#617) 
* fix(booking-templates): scope template list to the active company

GET /api/settings/booking-templates relied solely on the btl_select RLS
policy, which is membership-wide (user_company_ids) and returns templates
from every company the user belongs to. A user who owns multiple companies
saw all their templates merged regardless of which company was active.

Narrow the list in the API layer (mirroring counterparty-templates) to
system + the active company + the active company's team. RLS stays the
security backstop; this fixes the cross-company merge within a single
user's own view (it was never a cross-tenant data leak).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(import): show proper message for duplicate bank file upload

The bank file import page mis-parsed the structured error envelope
({ error: { code, message, details } }), so a BANK_FILE_DUPLICATE
(409) fell through to the generic "Kunde inte läsa filen" fallback.
The upload step also hardcoded that same string as the error heading,
so duplicates were doubly misreported as parse failures.

- Parse the structured envelope by error.code; surface error.message
  for all codes instead of rendering the error object.
- Add a dedicated BANK_FILE_DUPLICATE message using the importedAt /
  importedCount details the route already returns.
- Add an optional errorTitle prop to BankFileUploadStep (defaults to
  the previous text) and pass "Filen är redan importerad" for dupes.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(tests): add comprehensive tests for recordateEntry, inbox-linking, and external-id handling

- Implemented unit tests for recordateEntry in the bookkeeping module to validate various scenarios including date changes, non-posted entries, and fiscal period restrictions.
- Created tests for inbox-linking status in pending operations to ensure correct handling of invoice inbox items and supplier invoices, addressing historical bugs related to status updates.
- Added tests for external-id utilities to ensure consistent handling of monetary amounts and deduplication keys across different transaction sources.
- Introduced new functions in external-id.ts for stable external ID generation and normalization of imported descriptions, enhancing transaction deduplication reliability.

feat(migrations): add new database migrations for transaction handling

- Created migration to exclude storno and correction vouchers from unmatched GL lines, ensuring accurate reconciliation.
- Added a migration to preserve original bank transaction descriptions in a new immutable column, allowing for user edits while maintaining audit trails and deduplication integrity.

* feat(migrations): add function to exclude storno/correction vouchers from unmatched GL lines

* feat(transactions): enhance transaction handling with improved description normalization and preloaded original entries

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-01 14:45:49 +02:00
Jakob Wennberg 13be0c569a feat(mcp): expose multi-tx RPCs (match_batch_allocate + bulk_book_transactions) (#614) 
* feat(bulk-book): manual booking mode + document inheritance

Two pieces of user feedback from PR #606:

1. "How come it is only mallar? Is it not possible to have manuell
   bokfoering?" - BulkBookDialog was template-only. Added a Tabs
   primitive with Mall / Manuell tabs. Manual tab pre-fills lines from
   the selected txs (one line per tx on 1930 + counterparty
   placeholder on 3001/5800 by direction), then the user edits Konto /
   Debet / Kredit / Beskrivning. Live balance + bank-leg checks drive
   the confirm button - same invariants the RPC enforces server-side.

2. "Documents attached does not follow into the bookkeeping. And if
   there are two different documents attached, none of them follow."
   The bulk_book_transactions RPC now propagates each tx's document
   onto the target verifikat (new in Branch B, existing in Branch A)
   as verifikationsunderlag. Per BFL 5 kap 6§ + BFNAR 2013:2 kap 4 a
   verifikat may have multiple underlag; every receipt that justified
   a tx is now retention-protected on the combined entry. The dialog
   shows a small count chip ("N bilagor foeljer med") so the user
   sees what will inherit.

Also dropped p_user_id from the RPC signature (round-3 hardening
pattern applied consistently across all multi-tx RPCs after PR #607).
Caller resolves from auth.uid() inside the function.

Schema: BulkBookSchema is now a 3-way XOR
(existing_journal_entry_id | template_id+mode | manual_lines), with
manual_lines validated as accountNumber + nonNegativeAmount per line.

pg-real tests:
- doc inheritance into a new combined verifikat (mixed: 2 of 3 txs
  have docs - docs_linked should be 2, not 3)
- doc inheritance into an existing posted verifikat (link branch)
- manual lines path (no template expansion artifacts in the
  resulting JE - just the 2 user lines)
- unbalanced manual lines still rejected by BULK_BOOK_UNBALANCED

Migration applied to remote.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bulk-book): PR #610 review - pg-real signature, account allowlist, account-number validity

Three review findings on PR #610:

1. pg-real failure: 2 link-existing tests still used 5-arg SELECT
   bulk_book_transactions($1::uuid[], $2, $3, $4, $5) after the userId
   removal. My earlier replace_all caught only the patterns that had
   ::jsonb on $3; the link-existing tests pass null for new_entry and
   used a bare $3 so they slipped through. (Greptile P1)

2. Manual lines bypassed chart_of_accounts validation. A typo or
   adversarial caller could post to a BAS account that doesn't exist
   in this company's chart, corrupting the hauptbok and breaking SIE
   export. Both compliance-swarm (OWASP V2.3) and swedish-compliance
   flagged this. Added a single-roundtrip allowlist check in the
   route: query chart_of_accounts for distinct account_numbers in
   manual_lines and reject with BULK_BOOK_INVALID_ACCOUNT if any are
   missing or inactive.

3. UI canConfirm guard missed invalid account numbers. Account input
   allows 1-3 digits and JS string comparison '193' >= '1900' is false,
   so a 3-digit entry escapes bankLineNet, the bank match could pass
   via other lines, and the server returned 400 only after submit.
   Added previewLines.every(l => /^\d{4}$/.test(l.account_number)) to
   canConfirm so the Confirm button stays disabled inline.
   (Greptile P2)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bulk-book): PR #610 round 2 - RPC chart-of-accounts, doc tenant isolation, GRANTs

Seven compliance findings from the round-1 bot reviews:

Migration (20260602121000_bulk_book_round2_fixes.sql):
- RPC chart-of-accounts allowlist (defense-in-depth): every line in
  p_new_entry.lines is now verified to be an active BAS account for
  p_company_id. Closes the gap where the template branch and direct
  DB callers (psql, future MCP) bypassed the route's manual-branch
  check. Returns BULK_BOOK_INVALID_ACCOUNT with the offending list.
  (OWASP V8.2.1 + SOC 2 CC6.3)
- Document inheritance CTE: added "AND d.company_id = p_company_id"
  to the UPDATE join so the tenant isolation is enforced on both
  sides (tx + doc), not just the tx side. Four bots converged on this
  finding (V1.2.5, A.8.2, CC6.6, swedish-compliance).
- Bank-leg range check: "length(account_number) = 4 AND account_number
  BETWEEN '1900' AND '1999'" replaces the bare lexicographic comparison.
  Lexicographic-on-4-digit is safe today; the length guard is
  defense-in-depth against schema drift. (swedish-compliance)
- Explicit role grants: REVOKE ALL FROM PUBLIC + GRANT EXECUTE TO
  authenticated on both bulk_book_transactions and match_batch_allocate.
  (SOC 2 CC6.1)

UI (BulkBookDialog):
- Manual-mode prefill no longer suggests a hardcoded 3001/5800
  counterpart. Reason (swedish-compliance): a user accepting the
  prefill could submit a verifikat with no VAT line (26xx),
  under-reporting utgaaende moms. The bank side stays pre-filled
  (unambiguous); the counterpart row scaffolds blank for the user
  to choose.

Schema (BulkBookSchema):
- manual_lines.debit_amount + credit_amount bounded at 99,999,999 SEK
  per line. Catches typos before the RPC. (compliance-swarm V4.5)

i18n:
- docs_inherit_hint terminology: "bilaga" -> "verifikationsunderlag"
  and an explicit "sparas i 7 ar enligt BFL 7 kap" reminder.
  swedish-compliance flagged that "bilaga" risks users treating the
  files as deletable attachments rather than retention-bound
  raekenskapsinformation.

Migration applied to remote.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(test): seed chart_of_accounts in bulk-book pg-real seedTenant

The round-2 RPC fix added a chart_of_accounts allowlist check inside
bulk_book_transactions, but the test fixtures don't seed COA — so
every existing test that submits lines (1930, 3001, 2611, etc.) now
returns BULK_BOOK_INVALID_ACCOUNT instead of the expected error code.

Seed the 8 accounts the suite actually uses directly in seedTenant
(cheaper than calling seed_chart_of_accounts which inserts the full
BAS 2026 chart).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(mcp): expose match_batch_allocate + bulk_book_transactions as MCP tools

Surfaces the multi-tx flows shipped in PRs #603/#606/#608/#610 so
Claude Desktop/Code can drive them via chat.

- migration 20260603120000: expand pending_operations.operation_type
  CHECK to include match_batch_allocate, bulk_book_transactions, plus
  undo_sie_import (which was missing from prior expansions despite
  being wired in risk-tiers.ts and the commit dispatcher).
- types/index.ts: extend PendingOperationType.
- lib/pending-operations/risk-tiers.ts: match_batch_allocate = medium
  (same tier as single-tx match), bulk_book_transactions = high
  (creates a verifikat with arbitrary lines, same surface as
  create_voucher).
- lib/pending-operations/commit.ts: thin commit handlers that call
  the SQL RPCs and translate the structured error envelope. The RPCs
  themselves do all the locking, balance checks, JE creation, voucher
  number, payment/junction rows, and doc inheritance.
- extensions/general/mcp-server/server.ts: two new tool definitions.
  Both stage via stagePendingOperation with period_status hint and
  pre-validate inputs (direction, sum-equals-tx-abs, same-date,
  not-already-booked) so the agent gets a clear error inline before
  the RPC runs.
- payload-size.bench: bump from 30K to 31K tokens (with rationale).
  Two new tools earn the bump; descriptions already trimmed to fit
  the <=280-char description limit.

Migration applied to remote and version aligned with local filename.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): PR #614 review - allocation guard, IDOR pre-check, currency + JE-date

Round-1 review fixes on PR #614:

- Greptile P1: per-allocation invoice_id / supplier_invoice_id guard.
  The inputSchema marks both as optional (they're mutually exclusive
  by kind), so JSON Schema can't express "X required iff Y=A". Added
  explicit check in the execute handler: customer_invoice rows must
  carry invoice_id; supplier_invoice rows must carry supplier_invoice_id.

- OWASP V8.2.1: IDOR pre-check on match_batch_allocate. Verify every
  invoice / supplier_invoice referenced in the allocations belongs to
  this company BEFORE staging. The RPC re-checks (BATCH_INVOICE_NOT_FOUND),
  but failing fast at the MCP layer gives the agent a clear error.

- OWASP V8.2.1: same pre-check on bulk_book_transactions for
  existing_journal_entry_id. Fetches the JE at stage time, verifies
  status=posted and company_id, throws if not found.

- swedish-compliance: currency homogeneity check on bulk_book. Mixed
  SEK + EUR in one samlingsverifikat violates BFL 5 kap 6§ st 3 motpart
  clarity. Cross-currency batches go through match_batch_allocate
  instead (which handles FX diff on 7960/3960).

- swedish-compliance: period-lock check on the link-existing branch now
  uses MAX(tx_date, JE.entry_date), not just tx_date. Otherwise a tx in
  an open period could attach to a verifikat in a locked period and
  the guard would miss it.

- A.8.11 + CC7.2: sanitised RPC error logging. log.error now emits only
  { code, message } instead of the full error object — error.details can
  echo invoice IDs, amounts, and counterparty identifiers.

Not actioned (PR-comment, no code change):
- V2.3 double-validation in commit handler — RPC enforces balance,
  accounts, bank-leg via the chart_of_accounts allowlist (PR #610 round 2).
  Commit handler is a thin pass-through by design.
- A.8.2 step-up approval for high-tier ops — architectural change
  affecting all high-tier ops, not PR-scoped.
- V2.4 rate limiting on bulk endpoints — platform-level concern.
- 0.005 epsilon / account-class allowlist — pre-existing patterns.
- undo_sie_import storno requirement — separate RPC, this PR only
  backfilled the missing CHECK constraint.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): PR #614 round 2 - trust-boundary comments + balance pre-check + audit log

Round-2 review fixes (compliance-swarm went 14 -> 9 after round 1;
remaining HIGHs are all "do the same tenant check at multiple
layers"). The bot itself offers the alternative: "or document and
reference the specific RPC line that enforces this." Following that.

- commit.ts: trust-boundary comment blocks on both
  commitMatchBatchAllocate and commitBulkBookTransactions, citing the
  exact RPC + migration where tenant isolation + chart_of_accounts
  allowlist are enforced authoritatively. The commit handler stays a
  thin pass-through by design; re-querying would triple the same check
  without adding security. (V8.2.1, A.8.2)

- commit.ts: structured success-path log.info() on both handlers with
  companyId, operationType, journal_entry_id, and tx count. No raw
  amounts or IDs that could echo PII. (V16)

- server.ts: balance pre-check on bulk_book create-new path. RPC
  enforces BULK_BOOK_UNBALANCED authoritatively, but failing fast at
  staging gives the agent a clear error before pending_operations is
  even touched. (V2.3 / swedish-compliance)

Not actioned this round:
- V2.2 oneOf/if-then-else in JSON Schema for mutual exclusivity — JSON
  Schema vocabulary support is shaky across MCP clients; runtime check
  in execute() is the canonical pattern across the existing toolset.
- CC6.1 generic error string to caller — RPC error codes are
  user-actionable (BULK_BOOK_UNBALANCED, BATCH_INVOICE_NOT_FOUND); a
  generic string would degrade UX.
- CC7.2 audit RPC RAISE messages for PII — separate audit; not
  PR-scoped.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): PR #614 round 3 — last 5 LOWs + salary_run/agi constraint backfill

Compliance-swarm went 14 → 9 → 5 (all LOW). Cleaning the last 5 + the
swedish-compliance findings.

- migration 20260603121000: backfill create_salary_run + generate_agi
  into pending_operations.operation_type CHECK. Both have risk-tier
  entries and commit executors but were never added (same bug class
  as undo_sie_import). Production has no rows of either type today.
  (swedish-compliance)

- server.ts: Number.isFinite guard in bulk_book balance pre-check.
  Number(x) || 0 silently treats NaN as 0 — a malformed amount could
  pass the balance check by accident. (compliance-swarm A.8.28)

- server.ts: count-equality + missing-set assertion in match_batch_allocate
  tenant pre-check. Belt-and-suspenders so a null/undefined row in the
  Supabase JSON response can't pass silently. Same pattern on both
  invoice and supplier_invoice branches. (CC6.1)

- server.ts: fix BFL paragraph citation in currency-homogeneity
  comment. Was "BFL 5 kap 6§ st 3", should be "BFL 5 kap 2§" (SEK
  denomination) read with 5 kap 6§ (valutakurs). (swedish-compliance)

- server.ts: clarify 0.005 tolerance comment — it's for floating-point
  equalisation only, not a rounding allowance. RPC enforces exact
  balance to the öre. (swedish-compliance)

- commit.ts: expand audit-log txId comment — included intentionally
  for trail-to-source join, scoped to companyId already logged.
  (compliance-swarm A.8.15/CC7.2)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): PR #614 round 4 — Swedish plural typo + balance comment parity + agent-routing hint

Round-3 review caught:

- swedish-compliance: \`kundfakturaor\` typo (real räkenskapsinformation
  defect under BFL 5 kap 7§). Swedish plural for \`kundfaktura\` is
  \`kundfakturor\` (drop the final \`a\`, add \`or\`), same for
  \`leverantörsfaktura\` → \`leverantörsfakturor\`. Fixed via slice(-1) + 'or'.

- swarm A.8.28: match_batch_allocate balance tolerance check was
  missing the equivalent "RPC enforces exact balance" comment that
  bulk_book has. Added.

- swedish-compliance: currency-mismatch error message now routes the
  agent to gnubok_match_batch_allocate for cross-currency allocations
  instead of letting it retry with hand-built FX lines.

Not actioned (out of pattern / out of scope):
- Integer arithmetic for balance checks (codebase pattern is float +
  epsilon; would diverge from match_batch_allocate, supplier-payment,
  invoice-payment, etc.)
- DSD docs / runbook for txId-in-log and stripped-error.details
  trade-offs (out of PR scope; tracked separately)
- Link-existing target verifikat description match (architectural;
  every link-existing op would need this)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(mcp): expose link_transaction_to_journal_entry as MCP tool

The REST endpoint /api/transactions/[id]/link-journal-entry already lets the
duplicate-payment UI attach a bank tx to an already-posted verifikat without
creating new bookkeeping. Agents had no equivalent — closing that parity gap
so users on Claude can match bank txs against vouchers they booked manually.

The core link logic moves to lib/transactions/link-journal-entry.ts so both
the REST route and the new commit handler share one implementation (preserves
all structured-error codes, optimistic-lock invoice update, and compensating
rollback). New 'link_transaction_journal_entry' op type wired through the
risk tiers (medium), TOOL_SCOPE_MAP (transactions:write), and dispatcher.

Bumps the tools/list payload-size ceiling 31K → 31.5K — same family bump
PRs #603/#606 made when adding match_batch_allocate / bulk_book_transactions.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): PR #614 round 5 — bot findings on link_transaction_journal_entry

Addresses the swedish-compliance + compliance-swarm findings on commit 5b884c3a:

1. **CHECK constraint backfill** — new migration adding 'link_transaction_journal_entry'
   to pending_operations.operation_type. Same bug class as the salary_run/agi backfill
   in 20260603121000; without it, every staged op would be rejected silently in
   production (BFL 5 kap 6–7§ audit-trail gap).

2. **Payment-date exchange rate** — invoice_payments.exchange_rate now uses
   transaction.exchange_rate (rate on payment date) instead of invoice.exchange_rate
   (rate on invoice date), per BFL 5 kap 2§ + ML 8 kap 21–23§. The full 3960/7960
   posting still belongs to createInvoicePaymentJournalEntry by contract — this path
   only links to an EXISTING verifikat.

3. **voucherLabel format centralized** — exported formatVoucherLabel helper returns
   the canonical `A-12` format (with hyphen, matches gnubok_link_invoice_to_voucher
   and SIE #VER cross-references). Both the MCP staging preview and the committed
   service result import it, so the user can't approve one label and have a
   different one land in the audit trail.

4. **Rollback warn log restored** — txLog.warn-equivalent (IDs only, no PII) when
   the compensating rollback itself fails, surfacing partial-state gaps for
   reconciliation per GDPR Art.5(1)(f) / SOC 2 CC7.2. Lost in the refactor that
   extracted the shared service; now present in both rollback call sites.

5. **Commit-layer log.info** — structured success log mirroring
   commitMatchBatchAllocate / commitBulkBookTransactions (companyId, tx/JE IDs,
   settledInvoice boolean). No raw amounts or counterparty names.

6. **Data minimization on invoice fetch** — explicit column list replaces
   select('*, customer:customers(name)') in the shared service; the MCP staging
   pre-check now fetches only invoice_number + remaining_amount (drops total +
   paid_amount). voucher_description omitted from preview_data per Art.25.

Test impact: existing route + dispatcher tests updated to expect `A-12` instead
of `A12`. All 4308 tests pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(invoices): correct FX bookkeeping + UI for match-invoice flow

User report: matching a 230 SEK bank tx against a 140 USD invoice produced
1930 Dr 2 142,50 / 1510 Cr 2 142,50 — fictitious numbers that didn't match
either the bank receipt or the booked AR. Root cause: the preview route
called resolveSekAmount(tx.amount, null, INV.currency, INV.rate), treating
the SEK tx number as if it were in the invoice's currency and multiplying
by the invoice's stored rate. Both the preview and the commit then used
the bogus number on both legs and silently dropped the FX gain/loss.

A second issue surfaced in the same dialog: for a 1 250 SEK invoice with a
prior 230 SEK partial, the comparison row showed "Differens: 250 kr" (off
the original total) instead of "20 kr" (off the actual 1 020 kr remaining).

This patch:

1. **New shared helper** lib/bookkeeping/invoice-payment-lines.ts
   - buildInvoicePaymentClearingLines(tx, invoice, description) → bank-leg,
     AR-leg, fx-diff, and a balanced line array. Bank-leg is always the
     actual SEK that hit the bank (resolveSekAmount with the TX's currency
     context, honouring tx.amount_sek when set). AR-leg is the SEK value
     of the customer-debt reduction at the invoice's stored rate. Diff
     posts to 3960 (gain) or 7960 (loss) so the verifikat balances per
     BFL 5 kap 4–5§. Mirrors the match_batch_allocate RPC's contract:
     when the tx is cross-currency, the single match fully clears the
     invoice's remaining amount.

2. **Preview route** uses the helper for the clearing branch — replaces
   the buggy resolveSekAmount call. Now byte-identical to what commit
   builds.

3. **Match-invoice POST** uses the helper + createJournalEntry directly
   for the clearing path, bypassing createInvoicePaymentJournalEntry on
   this single flow. mark-paid and other callers of that function still
   work as before (full payment + caller-supplied exchangeRateDifference).

4. **InvoiceMatchDialog** compares the bank tx against
   invoice.remaining_amount (not invoice.total) for both customer and
   supplier branches; cross-currency dialogs now show the different-
   currencies warning instead of a meaningless numeric diff. The dialog's
   invoice card also displays remaining_amount.

8 new unit tests cover same-currency full/partial, cross-currency gain/loss,
exact match (no FX line), sub-öre tolerance, and USD-on-USD with pre-
populated amount_sek. All 4316 tests pass.

Scope note: this expands PR #614 beyond the original "expose multi-tx RPCs
as MCP tools" since the same FX bug class affected the new MCP tool too
(round 5 already addressed the invoice_payments.exchange_rate side).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): PR #614 round 7 — CI build + 4 HIGH bot findings

Core Build was failing on e29a0ba2/5e9d4c3d due to a TypeScript type-cast
error in linkTransactionToJournalEntry. Plus the swedish-compliance review
flagged four substantive bugs in my recent commits.

1. **TS build error** — `invoice = invoiceRow as typeof invoice` inferred
   `never` because the LHS type included `null`. Switched to a named
   `FetchedInvoice` alias and `as unknown as FetchedInvoice`.

2. **TOOL_SCOPE_MAP missing two write-capable tools** (🟠 HIGH OWASP V8.2.1).
   `gnubok_match_batch_allocate` and `gnubok_bulk_book_transactions` (added
   in PRs #603/#606) were never registered, meaning any API key could invoke
   them regardless of scope. Backfilled both with `transactions:write`.

3. **`paymentExchangeRate` fallback wrong-date rate** (swedish-compliance).
   `transaction.exchange_rate ?? invoice.exchange_rate ?? null` falls back
   to the INVOICE date's rate when the tx rate is null. Per ML 8 kap 21–23§
   the payment row must record the PAYMENT-date rate. Removed the fallback
   — `null` is correct when the tx is SEK; downstream lookups can populate
   it lazily from Riksbanken if needed.

4. **Currency-mismatch corrupts paid_amount** (swedish-compliance). The
   link path was accumulating `tx.amount` into `invoice.paid_amount` without
   checking that the currencies matched. A 230 SEK tx applied to a USD
   invoice would record "230 USD paid" silently. Added explicit
   LINK_TX_INVOICE_CURRENCY_MISMATCH guard (400) — cross-currency
   settlement must go through the match-invoice flow which routes through
   buildInvoicePaymentClearingLines.

5. **Cross-currency PARTIAL overstates FX gain/loss** (swedish-compliance,
   BFL 5 kap 4–5§). `buildInvoicePaymentClearingLines` was crediting the
   FULL invoice remaining to 1510 on every cross-currency match — zeroing
   the GL balance while the invoice row stayed at status=partially_paid,
   and booking a fake huge FX diff to 3960/7960. Fix: only book FX-diff
   when `bankSek >= arSekFullRemaining`. Partials default to
   1930 = 1510 = bankSek, deferring the FX adjustment to the final
   settlement (or to a manual mark-paid with explicit
   exchange_rate_difference). Documented the helper as customer-invoice-
   only (supplier-side has different DR/CR polarity and goes through
   match_batch_allocate RPC).

Test impact: 1 helper test updated to match the defer-on-ambiguous-loss
behavior, 1 new test covers the partial-defers-FX path explicitly. All
4317 tests pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): PR #614 round 8 — close out remaining bot findings

CI green on round 7 (4 of 4 checks), HIGH count 2 → 1. Round-8 closes
the remaining HIGH and the smaller doc/guard items.

1. **PI1.3 risk acknowledgment restored** (SOC 2 HIGH). The shared
   rollbackTxLink helper already had warn-level logging on rollback
   failure, but the explicit PI1.3 reference comment from the original
   route was lost in the refactor. Added inline so the reconciliation-
   gap risk is visible to future maintainers.

2. **MCP currency-mismatch pre-stage check.** gnubok_link_transaction_to_
   journal_entry now fetches invoice.currency and rejects cross-currency
   matches before staging, saving the user an approval round-trip when
   the commit handler's LINK_TX_INVOICE_CURRENCY_MISMATCH guard would
   fire anyway.

3. **fxDiffSek JSDoc clarified.** The sign convention (positive = loss,
   negative = gain) is correct for verifikat balancing but counter-
   intuitive at a P&L glance. Documented explicitly + pointed callers
   needing a "gain" number at `bankSek - arSek`.

4. **Reject both invoice_id + supplier_invoice_id** on the same
   match_batch_allocate row (V4.5). Extra IDs previously leaked into
   preview_data silently.

5. **Reject zero-amount tx** in bulk_book_transactions direction guard
   (A.8.28). A txs[0].amount === 0 would have mis-classified the batch
   as 'expense'. Mirrors the existing guard in match_batch_allocate.

6. **Reject debit=0 && credit=0 lines** in bulk_book new_entry (BFL 5
   kap 6§ — every verifikat line must represent a real bokföringspost
   with a non-zero amount).

7. **Data-minimization comments** added on the match-invoice preview
   route (amount_sek + exchange_rate fetch is for the FX-fix bank-leg
   math) and on the bulk_book_transactions preview_data block (aggregate
   counts only — no per-tx PII). Mirrors the pattern already documented
   on gnubok_link_transaction_to_journal_entry.

Skipped:
- 1510 vs 1515 (osäkra kundfordringar) — future improvement, needs
  reading the original invoice JE's account, not a single-tool fix.
- transaction_description PII masking in preview_data — needs product
  call on the truncation strategy and would degrade approval-UX.
- "invoice.match_confirmed event removed" finding — false positive; the
  event is emitted at lib/transactions/link-journal-entry.ts:270-280.

All 4317 tests pass; payload-size guard still under ceiling.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(invoices): PR #614 round 9 — block cross-currency in single match-invoice path

Closes the swedish-compliance finding from round-8 review: a SEK bank tx
matched against a USD invoice through /api/transactions/[id]/match-invoice
would silently corrupt invoice.paid_amount (accumulator treats SEK as USD)
and flip a 140 USD invoice to status='paid' after a tiny partial. The
round-6/7 FX fix corrected the JOURNAL ENTRY lines but the invoice STATE
update still ran the same broken accumulator.

Proper cross-currency settlement on this path requires converting tx.amount
to invoice.currency at the bank-date rate AND storing invoice_payments rows
with the right (amount, currency) pair. That's a larger design call that
belongs in its own PR.

This change blocks cross-currency on the single-allocation path:
- New MATCH_INVOICE_CURRENCY_MISMATCH structured error (400, bilingual)
- Same-currency check inserted right after MATCH_INVOICE_NOT_OPEN
- Mirrors the LINK_TX_INVOICE_CURRENCY_MISMATCH guard added to the link
  path in round-7
- Routes the user to the multi-allocation flow (gnubok_match_batch_allocate)
  which DOES handle 3960/7960 FX-diff postings end-to-end

Same-currency (SEK→SEK or USD→USD) remains fully supported including
partials; the buildInvoicePaymentClearingLines helper handles those correctly.
For SEK tx → USD invoice the user now gets a clean 400 error pointing at
the right flow, instead of silently corrupted ledger state.

1 new route test covers the guard. All 4318 tests pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-30 13:36:46 +02:00
Jakob Wennberg 28f7cefc86 feat(bulk-book): manual booking mode + document inheritance (#610) 
* feat(bulk-book): manual booking mode + document inheritance

Two pieces of user feedback from PR #606:

1. "How come it is only mallar? Is it not possible to have manuell
   bokfoering?" - BulkBookDialog was template-only. Added a Tabs
   primitive with Mall / Manuell tabs. Manual tab pre-fills lines from
   the selected txs (one line per tx on 1930 + counterparty
   placeholder on 3001/5800 by direction), then the user edits Konto /
   Debet / Kredit / Beskrivning. Live balance + bank-leg checks drive
   the confirm button - same invariants the RPC enforces server-side.

2. "Documents attached does not follow into the bookkeeping. And if
   there are two different documents attached, none of them follow."
   The bulk_book_transactions RPC now propagates each tx's document
   onto the target verifikat (new in Branch B, existing in Branch A)
   as verifikationsunderlag. Per BFL 5 kap 6§ + BFNAR 2013:2 kap 4 a
   verifikat may have multiple underlag; every receipt that justified
   a tx is now retention-protected on the combined entry. The dialog
   shows a small count chip ("N bilagor foeljer med") so the user
   sees what will inherit.

Also dropped p_user_id from the RPC signature (round-3 hardening
pattern applied consistently across all multi-tx RPCs after PR #607).
Caller resolves from auth.uid() inside the function.

Schema: BulkBookSchema is now a 3-way XOR
(existing_journal_entry_id | template_id+mode | manual_lines), with
manual_lines validated as accountNumber + nonNegativeAmount per line.

pg-real tests:
- doc inheritance into a new combined verifikat (mixed: 2 of 3 txs
  have docs - docs_linked should be 2, not 3)
- doc inheritance into an existing posted verifikat (link branch)
- manual lines path (no template expansion artifacts in the
  resulting JE - just the 2 user lines)
- unbalanced manual lines still rejected by BULK_BOOK_UNBALANCED

Migration applied to remote.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bulk-book): PR #610 review - pg-real signature, account allowlist, account-number validity

Three review findings on PR #610:

1. pg-real failure: 2 link-existing tests still used 5-arg SELECT
   bulk_book_transactions($1::uuid[], $2, $3, $4, $5) after the userId
   removal. My earlier replace_all caught only the patterns that had
   ::jsonb on $3; the link-existing tests pass null for new_entry and
   used a bare $3 so they slipped through. (Greptile P1)

2. Manual lines bypassed chart_of_accounts validation. A typo or
   adversarial caller could post to a BAS account that doesn't exist
   in this company's chart, corrupting the hauptbok and breaking SIE
   export. Both compliance-swarm (OWASP V2.3) and swedish-compliance
   flagged this. Added a single-roundtrip allowlist check in the
   route: query chart_of_accounts for distinct account_numbers in
   manual_lines and reject with BULK_BOOK_INVALID_ACCOUNT if any are
   missing or inactive.

3. UI canConfirm guard missed invalid account numbers. Account input
   allows 1-3 digits and JS string comparison '193' >= '1900' is false,
   so a 3-digit entry escapes bankLineNet, the bank match could pass
   via other lines, and the server returned 400 only after submit.
   Added previewLines.every(l => /^\d{4}$/.test(l.account_number)) to
   canConfirm so the Confirm button stays disabled inline.
   (Greptile P2)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bulk-book): PR #610 round 2 - RPC chart-of-accounts, doc tenant isolation, GRANTs

Seven compliance findings from the round-1 bot reviews:

Migration (20260602121000_bulk_book_round2_fixes.sql):
- RPC chart-of-accounts allowlist (defense-in-depth): every line in
  p_new_entry.lines is now verified to be an active BAS account for
  p_company_id. Closes the gap where the template branch and direct
  DB callers (psql, future MCP) bypassed the route's manual-branch
  check. Returns BULK_BOOK_INVALID_ACCOUNT with the offending list.
  (OWASP V8.2.1 + SOC 2 CC6.3)
- Document inheritance CTE: added "AND d.company_id = p_company_id"
  to the UPDATE join so the tenant isolation is enforced on both
  sides (tx + doc), not just the tx side. Four bots converged on this
  finding (V1.2.5, A.8.2, CC6.6, swedish-compliance).
- Bank-leg range check: "length(account_number) = 4 AND account_number
  BETWEEN '1900' AND '1999'" replaces the bare lexicographic comparison.
  Lexicographic-on-4-digit is safe today; the length guard is
  defense-in-depth against schema drift. (swedish-compliance)
- Explicit role grants: REVOKE ALL FROM PUBLIC + GRANT EXECUTE TO
  authenticated on both bulk_book_transactions and match_batch_allocate.
  (SOC 2 CC6.1)

UI (BulkBookDialog):
- Manual-mode prefill no longer suggests a hardcoded 3001/5800
  counterpart. Reason (swedish-compliance): a user accepting the
  prefill could submit a verifikat with no VAT line (26xx),
  under-reporting utgaaende moms. The bank side stays pre-filled
  (unambiguous); the counterpart row scaffolds blank for the user
  to choose.

Schema (BulkBookSchema):
- manual_lines.debit_amount + credit_amount bounded at 99,999,999 SEK
  per line. Catches typos before the RPC. (compliance-swarm V4.5)

i18n:
- docs_inherit_hint terminology: "bilaga" -> "verifikationsunderlag"
  and an explicit "sparas i 7 ar enligt BFL 7 kap" reminder.
  swedish-compliance flagged that "bilaga" risks users treating the
  files as deletable attachments rather than retention-bound
  raekenskapsinformation.

Migration applied to remote.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(test): seed chart_of_accounts in bulk-book pg-real seedTenant

The round-2 RPC fix added a chart_of_accounts allowlist check inside
bulk_book_transactions, but the test fixtures don't seed COA — so
every existing test that submits lines (1930, 3001, 2611, etc.) now
returns BULK_BOOK_INVALID_ACCOUNT instead of the expected error code.

Seed the 8 accounts the suite actually uses directly in seedTenant
(cheaper than calling seed_chart_of_accounts which inserts the full
BAS 2026 chart).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-30 10:23:14 +02:00
Mattsson ea1bf01f1e Fix/m sprint fixes (#613) 
* fix(dashboard): exclude ignored and already-triaged transactions from stale count

The "Gamla transaktioner" widget counted transactions that had been ignored
or already marked as is_business=true but not yet booked, so users saw a
nag for a row they had already dealt with — and the /transactions inbox
correctly hid it. Align the count with the inbox criterion (is_business
IS NULL, is_ignored = false) so the widget clears when the row leaves
the inbox.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(transactions): read entity_type from settings response wrapper

The transactions page read entityRes.entity_type directly, but
/api/settings returns { data: { entity_type, ... } }. The expression
was always undefined, so setEntityType never fired and entityType
stayed at its initial 'enskild_firma'. The template picker's
entity_type filter then dropped every aktiebolag-tagged user template
for AB customers — only entity_type='all' templates made it through.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* stale templates
bank sync
journal entry from transaction

* fixed pr comments

* fixed pr comment

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-30 01:28:41 +02:00
Jakob Wennberg fc7a46c3f2 fix(match-batch): cross-currency allocations + widened tolerance (#607) 
* fix(match-batch): cross-currency allocations + widened tolerance

Reported by jakob testing PR #603's MatchAllocationDialog with a SEK
bank tx + a mix of SEK and USD invoices:

1. Tally rendered "1 USD + 1 SEK = 2 kr" — summing different currencies
   as if they were the same.
2. The 0.005 SEK tolerance blocked confirm on any FX rounding delta.

## What changed

**UI (MatchAllocationDialog.tsx)**
- Per-row amount input is explicitly in TRANSACTION currency (SEK for a
  Swedish bank import). Cross-currency rows show an "≈ X.XX (invoice
  currency)" hint under the input so the user can verify the FX result.
- Default amount for a cross-currency allocation is
  `invoice.remaining × invoice.exchange_rate` (booked SEK), so the user
  doesn't have to mental-math the FX.
- Overshoot tolerance widened from 0.005 SEK to `max(1 SEK, 0.5% × tx)`
  so bank-side FX rounding doesn't block confirm. A 2 400 kr tx now
  accepts ~12 kr of tolerance, a 100 kkr transfer accepts 500 kr.

**RPC (match_batch_allocate cross_currency migration)**
- BATCH_CURRENCY_MISMATCH dropped per-allocation. Mixed currencies now
  accepted with the convention that the cross-currency row pays the
  FULL invoice remaining (matches the single-tx match-supplier-invoice
  behavior). Partial cross-currency is out of scope for v1.
- AR/AP line is booked at `invoice.remaining × invoice.exchange_rate`
  (the SEK that was originally on 1510/2440). FX residual is posted
  to 7960 (Valutakursförluster) or 3960 (Valutakursvinster) per BAS.
- Sign conventions per direction documented inline:
    Customer: bank > booked → Cr 3960 (gain); bank < booked → Dr 7960
    Supplier: bank < booked → Cr 3960 (gain); bank > booked → Dr 7960
- New BATCH_FX_RATE_MISSING when the cross-currency invoice has no
  exchange_rate on file (would otherwise silently book at 0).
- New BATCH_FX_DEVIATION_TOO_LARGE when the user-entered amount
  deviates more than 10% from booked SEK — catches typos like "140"
  (USD invoice currency) when they meant "1390" (SEK equivalent)
  without rejecting genuine rate-day FX movement.

RPC patched on remote via Supabase MCP. Same-currency path is
byte-identical to the previous behavior.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(match-batch): PR review — strict sum, bank line = tx_abs, FX validation

Round-1 review fixes on the cross-currency batch allocation flow:

UI (MatchAllocationDialog):
- Tighten tolerance to 0.005 SEK so the "balanced ✓" indicator matches
  what the server will accept. The previous widened tolerance (max 1 SEK
  or 0.5% × tx) created a reconciliation gap where the JE's bank line
  could legitimately disagree with the actual bank receipt.
- Require balanced before confirm — undershoot is now a blocking state
  with an explicit warning, not a silent "leave unallocated".
- Cross-currency default no longer caps at remainingTxBudget. Capping a
  USD invoice's default to the leftover SEK budget could silently
  trigger BATCH_FX_DEVIATION_TOO_LARGE on submit. The user re-balances
  the other rows to fit.
- Add explicit FX-rate validation (bound check 0 < rate < 100000).
- When a cross-currency invoice has no usable exchange_rate on file,
  leave the amount blank and surface a warning instead of guessing.

RPC (match_batch_allocate):
- New code BATCH_AMOUNT_BELOW_TX. Strict sum check on both sides means
  the server can't be coaxed by a direct API caller into the same
  broken state the UI now blocks.
- Bank line credit/debit = v_tx_abs (the actual bank movement) instead
  of sum-of-allocations. Same value within rounding under the strict
  sum check, but it makes intent legible and lets per-row FX diff lines
  absorb rounding.
- Defense-in-depth company_id filter on all re-queries / UPDATEs in
  the line-build + payment-row passes.
- Drop the v_booked_sek-aliasing-for-invoice.total foot-gun. Use a
  dedicated v_inv_total var.
- Truncate invoice_number to 32 chars in line_description.

Tests:
- pg-real: cross-currency happy path (USD invoice paid by SEK tx with
  FX loss to 7960, bank line = tx_abs).
- pg-real: BATCH_AMOUNT_BELOW_TX rejection on undershoot.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(match-batch): PR review round 2 - caller user_id verification + FX bound

Compliance-swarm + swedish-compliance findings on round 1:

- CC6.3 (HIGH): p_user_id was caller-supplied and written into
  journal_entries.user_id / payment-row user_id without verifying it
  equals auth.uid(). Membership covered the company; nothing covered
  the user attribution. Two-layer fix: explicit guard rejects when
  p_user_id <> auth.uid(), and all writes now resolve v_caller =
  auth.uid() directly so the guard cant be silently bypassed.
- A.8.28 (MED): server-side FX upper-bound (0 < rate < 100000) matches
  the UI. Previously RPC only checked > 0, allowing the UI guard to
  diverge.
- V1.2.5 (LOW): truncate v_tx.date when concatenated into
  line_description (defense alongside round 1s invoice_number trunc).
- Symmetry: populate supplier_invoice_payments.exchange_rate (column
  existed, INSERT omitted it). Customer side already populated. Matches
  swedish-compliances traceability note on AP rorelseskulder.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(match-batch): PR review round 3 - drop p_user_id, CHECK constraints, payment-day rate

Genuine round-2 review findings (compliance-swarm + swedish-compliance):

- V4.5: p_user_id dropped from RPC signature entirely. Round-2 added a
  guard; this removes the attack surface at the API boundary. Caller is
  resolved via auth.uid() inside the function. Route updated.
- V2.2: CHECK constraint on invoices.exchange_rate and
  supplier_invoices.exchange_rate (0 < rate < 100000). Three layers now
  enforce the bound: schema, RPC, UI.
- swedish-compliance traceability gap: payment_exchange_rate column on
  both invoice_payments and supplier_invoice_payments. Populated as
  v_alloc_amount / v_inv_remaining for cross-currency rows so FX diffs
  are reconstructible from the payment record alone (BFL 7 kap
  behandlingshistorik). NULL for same-currency. The existing
  exchange_rate column continues to store the invoicing rate.
- CC6.1: extract isValidExchangeRate() to lib/utils.ts. UI's three
  inline bound checks now share one validator.
- Dead code: drop unused leftover_note i18n key (sv + en).

Tests:
- pg-real signature updated (4-arg -> 3-arg) across all 9 call sites.
- Added payment_exchange_rate assertion to cross-currency happy path
  (invoicing rate 10.0 stays, payment-day rate stored as 10.5).

Migration applied to remote.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(test): missed 4th arg in BATCH_UNAUTHORIZED pg-real test

Round-3 dropped p_user_id from match_batch_allocate. The replace_all
caught the userId/companyId pattern but missed the BATCH_UNAUTHORIZED
test which uses outsiderId instead of userId. CI failed with
"bind message supplies 4 parameters, but prepared statement requires 3".

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-29 16:30:42 +02:00
Jakob Wennberg 4da87e5e4c feat(transactions): bulk-book + is-booked predicate (#606) 
* feat(transactions): bulk-book + is-booked predicate

Closes the second of the two multi-tx ↔ multi-voucher flows from the
original plan. Where PR #603's match_batch_allocate took 1 tx and
spread it across N invoices (samlingsbetalning), this PR takes N bank
transactions on the same day and rolls them up into ONE combined
verifikat (samlingsverifikation per BFL 5 kap 6§ st 3) — the kiosk
masshantering pattern the user explicitly asked for.

## Backend (Phase 3b)

- **PL/pgSQL RPC** bulk_book_transactions: two branches, both atomic.

  1. Link to existing posted verifikat (p_existing_journal_entry_id):
     no new JE. Validates the JE's 19xx net equals sum(tx.amount),
     inserts N transaction_voucher_links rows, and for N=1 also sets
     transactions.journal_entry_id (1:1 reader-path back-compat).

  2. Create new combined verifikat (p_new_entry with pre-computed
     balanced lines): the route's applyTemplate() has already done
     ratio + VAT expansion per the chosen mode. The RPC validates the
     lines balance and the 1930 net matches sum(tx.amount), then
     commits via commit_journal_entry.

  Same security pattern as match_batch_allocate: company-member check
  via auth.uid(), SELECT … FOR UPDATE on each tx in id order,
  deterministic fiscal-period resolution (ORDER BY period_start DESC).

- **Endpoint** POST /api/transactions/bulk-book — fetches template via
  RLS, expands per mode (one_line_per_tx | sum_per_account) using
  lib/bookkeeping/template-library.applyTemplate, passes the resulting
  lines to the RPC. On success emits one transaction.reconciled event
  per tx.

- **22 new BULK_BOOK_* error codes** (sv + en) covering all guard
  paths.

## UI (Phase 5b)

- **BulkBookDialog** — template picker + mode toggle (segmented
  control: en rad per transaktion / summera per konto) + live preview
  table with balance + bank-leg invariant indicators. Confirm only
  enabled when both pass.

- **Multi-select inbox** — sticky action bar gains a "Bokför i klump"
  button gated by same-date + same-direction across selected txs.
  Tooltip explains the disabled state.

## Phase 6: is-booked predicate

New lib/transactions/is-booked.ts. After multi-allocation and bulk-
book, tx.journal_entry_id can be NULL even though the tx is anchored
(via invoice_payments / supplier_invoice_payments /
transaction_voucher_links). The helper checks all three storage
locations so future readers don't falsely show multi-anchored txs as
"unbooked". Companion getPrimaryJournalEntryId() resolves the best
JE link to surface in UI. SQL mirror is_transaction_booked() exists
from the PR #602 foundation migration.

Existing readers (TransactionHistoryList, TransactionInboxCard) are
not yet refactored to use the helper — that's a follow-up that
touches per-tx JE links across multiple call sites. The helper is
documented + tested so subsequent refactors are mechanical.

## Tests

- tests/pg/bulk-book-transactions.pg.test.ts — 8 pg-real scenarios
  (happy path create-new with 3 txs, happy path link-existing, date
  mismatch, direction mismatch, amount mismatch, unbalanced lines,
  unauthorized).
- app/api/transactions/bulk-book/__tests__/route.test.ts — 5 unit
  tests (schema XOR, link path, create-new with template fetch +
  applyTemplate, structured-error mapping).
- lib/transactions/__tests__/is-booked.test.ts — 11 cases covering
  all three storage locations + primary-JE resolution.

26 unit tests pass on touched paths. RPC migration applied to remote
via Supabase MCP.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bulk-book): PR #606 review round 1 + CI fixes

Closes the build failure and the two real Greptile findings.

## CI

- **core-only + Vercel build fail**: I used useMemo for
  selectedTransactions and bulkBookEligible on the transactions page
  without importing it. TypeScript build (`next build`) caught it
  with "Cannot find name 'useMemo'". Fixed the import.

## Review findings

- **(P1) Currency mismatch returned BULK_BOOK_DIRECTION_MISMATCH**
  whose user-facing message blames direction. Mixed SEK + EUR batches
  would show "All transactions must be the same direction" which is
  factually wrong. Introduced dedicated BULK_BOOK_MIXED_CURRENCY code
  (sv + en) explaining the actual constraint, and switched the route
  to use it.

- **(P1) Branch B (create-new) N=1 missed
  reconciliation_method='manual'**. Branch A's N=1 UPDATE sets it
  alongside journal_entry_id; Branch B's didn't, leaving the
  reconciliation_method NULL even though the single tx was reconciled
  via the same flow. Downstream readers (reconciliation reports,
  status indicators) would treat the two N=1 paths differently. New
  follow-up migration patches Branch B's final UPDATE.

RPC patch applied to remote via Supabase MCP. 26 unit tests pass.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-29 14:55:24 +02:00
Jakob Wennberg 7eb8715417 feat(transactions): split-payment allocator — 1 tx → N invoices (#603) 
* fix(category-mapping): use leaf BAS accounts instead of group codes

3900, 5800, 6200 are BAS gruppkonton (header codes) and shouldn't carry
postings. Switched the default mappings to the matching leaf accounts:

  - income_other:     3900 -> 3999 (Övriga rörelseintäkter)
  - expense_travel:   5800 -> 5890 (Övriga resekostnader)
  - expense_telecom:  6200 -> 6230 (Datakommunikation)

The fallback for income_other inside getCategoryAccountMapping was also
hardcoded to '3900'; updated to '3999' for consistency.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(transactions): split-payment allocator — 1 tx → N invoices

Closes one of the two flows that motivated PR #602's foundation:
allocating a single bank transaction across multiple customer OR
multiple supplier invoices, with one combined verifikat
(samlingsverifikation per BFL 5 kap 6§ st 3).

## Backend (Phase 3a)

- **PL/pgSQL RPC** match_batch_allocate (~400 lines): locks the tx +
  each target invoice with SELECT … FOR UPDATE in id order, validates
  status/currency/remaining/direction before any write, builds the
  combined verifikat via commit_journal_entry (atomically assigns
  voucher_number + flips draft→posted), inserts N rows in
  invoice_payments or supplier_invoice_payments pointing at the same
  JE, advances paid_amount/remaining_amount/status per invoice. Returns
  { ok, journal_entry_id, voucher_number, allocations: [...] } on
  success or { ok: false, code, details } on guard failure. Mixed
  customer+supplier kinds are rejected (v1 scope).

- **Endpoint** POST /api/transactions/[id]/match-batch — thin wrapper
  around the RPC. Validates body via MatchBatchSchema (zod
  discriminatedUnion + superRefine to catch mixed-kinds at the schema
  layer). On RPC success, emits one invoice.match_confirmed or
  supplier_invoice.match_confirmed event per allocation so existing
  subscribers (reminders, automations, processing-history) keep
  working. Maps the structured RPC error envelope to
  errorResponseFromCode.

- **16 new BATCH_* error codes** (sv+en): BATCH_TX_NOT_FOUND,
  BATCH_TX_ALREADY_BOOKED, BATCH_OVERSHOOT, BATCH_AMOUNT_EXCEEDS_TX,
  BATCH_MIXED_KINDS_UNSUPPORTED, BATCH_DIRECTION_MISMATCH,
  BATCH_CURRENCY_MISMATCH, BATCH_PERIOD_LOCKED, BATCH_RPC_FAILED, etc.

## UI (Phase 5a)

- **MatchAllocationDialog** (components/transactions/) — direction-
  aware (positive tx → customer invoices, negative → supplier). Search
  + selectable list of open invoices. Per-row amount input with default
  = min(invoice.remaining, tx_remaining_budget). Live tally with
  green-check balanced state, red overshoot warning, gray leftover
  note. Confirm button disabled on overshoot. POSTs to /match-batch
  and on 200 triggers the same exit animation as single-tx match.

- **Inbox row** gains a second outline icon button (Split icon) next
  to the existing 1:1 match button, gated by the same
  showInvoiceMatchButton predicate. Tooltip explains the direction-
  aware split. Opens MatchAllocationDialog.

- **i18n** strings under tx_match_allocation namespace in sv.json
  and en.json (32 keys each).

## Tests

- tests/pg/match-batch-allocate.pg.test.ts — 5 pg-real tests covering
  combined verifikat shape, overshoot guard, already-booked tx,
  direction mismatch, mixed-kinds rejection.
- app/api/transactions/[id]/match-batch/__tests__/route.test.ts — 5
  unit tests covering schema validation, mixed-kinds, happy path,
  structured-error mapping, raw-error → BATCH_RPC_FAILED.

63 unit tests pass across the touched paths. The RPC migration was
already applied to remote in an earlier Phase 3a session (idempotent
CREATE OR REPLACE FUNCTION; the next replay is a no-op).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(match-batch): PR #603 review round 1 + CI fixes

Closes both CI failures and the three real review findings.

## CI fixes

- **pg-real failure**: the RPC declared
  `v_journal_entry_id uuid := uuid_generate_v4()` which fails in the CI
  Postgres image (uuid-ossp extension is off). Switched to
  `gen_random_uuid()` — the codebase standard already used by
  supplier_invoices, invoice_inbox, etc.
- **core-only failure**: my earlier BAS leaf-account commit
  (3900→3999, 5800→5890, 6200→6230) didn't update the matching
  `lib/bookkeeping/__tests__/category-mapping.test.ts` expectations,
  and `getDefaultAccountForCategory`'s fallback for `income_*` was
  still hardcoded to '3900'. Updated both.

## Review findings (greptile)

- **P1 deadlock-stable locking** (`match_batch_allocate.sql:11`): the
  validation `FOR UPDATE` loop ran in caller-supplied array order. Two
  concurrent calls with overlapping invoice sets in opposite orders
  could deadlock and one would abort with `BATCH_RPC_FAILED`. Now
  all three loops (validate, build lines, advance invoices) iterate
  via `SELECT … FROM jsonb_array_elements(…) ORDER BY
  COALESCE(invoice_id, supplier_invoice_id)`, giving a stable global
  lock order regardless of how the caller ordered the JSON array.

- **P1 duplicate-allocation detection** (`match_batch_allocate.sql:163`):
  the same invoice_id listed twice would pass the per-row overshoot
  guard (both iterations read the original `remaining_amount`) and
  the write loop would insert two `invoice_payments` rows for the
  same invoice. Added a `v_seen_ids text[]` check in the validation
  loop and a new `BATCH_DUPLICATE_ALLOCATION` error code (sv + en).
  The dialog already prevents this UI-side via `if (prev[candidate.id]
  return prev` — the RPC guard is the defense-in-depth layer.

- **P2 zod `.positive()`** (`schemas.ts:544`): allocation amount was
  `nonNegativeAmount` (allowing 0), passing schema validation only to
  be rejected by the RPC with `BATCH_INVALID_AMOUNT`. Now
  `z.number().positive(…)` so 0-amount entries fail at the schema
  layer with a per-field path, cleaner 400.

- **P2 strict `> 0` direction check** (`MatchAllocationDialog.tsx:82`):
  used `amount >= 0` to pick customer-side, but a zero-amount tx would
  load customer candidates only to hit `BATCH_TX_ZERO_AMOUNT` at
  submit time after the user has filled in allocations. Switched to
  `> 0` so 0-amount tx never reaches the dialog at all (it's rejected
  by the RPC immediately).

The fourth Greptile comment (the schema P2 about amount validation)
overlaps with the third; addressed in the same edit.

## Verification

- 112 unit tests pass across touched paths
- ESLint clean
- New pg-real test `tests/pg/match-batch-allocate.pg.test.ts` covers
  the dedupe scenario (same supplier invoice listed twice with summing
  amounts that individually pass per-row overshoot)
- RPC patch applied to remote via Supabase MCP

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(match-batch): PR #603 review round 2 — compliance hardening

Addresses the actionable findings from compliance-swarm and
Swedish-accounting-compliance reviews. Six small RPC changes + two
TS-side guards, all bundled in one follow-up migration.

## Security

- **(GDPR Art.5(1)(f) / ISO A.8.2) Caller verification**: SECURITY
  DEFINER bypasses RLS, and the prior RPC accepted any
  (p_user_id, p_company_id) pair from the route. Now the function
  rejects with new `BATCH_UNAUTHORIZED` (sv+en, HTTP 403) if
  `auth.uid()` is not a member of `p_company_id`. Pattern lifted from
  `harden_invoice_number_rpcs` (#20260510140000).
- **(OWASP V4.2) Allocation cap**: `MatchBatchSchema.allocations` now
  carries `.max(100)` to prevent DoS via unbounded FOR UPDATE locks.

## Swedish accounting correctness

- **source_type per direction**: was hardcoded to `'invoice_paid'` for
  both customer + supplier batches, mis-routing behandlingshistorik
  filters. Customer batches keep `'invoice_paid'`, supplier batches now
  write `'supplier_invoice_paid'`.
- **Fiscal-period determinism**: `LIMIT 1` on the period lookup was
  non-deterministic on overlap (e.g. corrected broken year). Added
  `ORDER BY period_start DESC` so the most recent matching period
  wins.
- **Tolerance harmonisation**: cross-allocation sum used `+0.01`
  tolerance while per-row used `+0.005`. Both now `+0.005` so a
  multi-row batch can't drift ~0.01 SEK while each row passes
  individually.
- **`transactions.category` no longer overwritten**: was forced to
  `'income_services'` (→ BAS 3001 at 25% VAT) for any customer batch,
  misrepresenting reduced-rate / export / EU-service invoices. The
  category is only meaningful 1:1 with a single invoice; batches now
  leave it as-is, mirroring the supplier-side `ELSE category` branch.

## Tests

- `tests/pg/match-batch-allocate.pg.test.ts` now wraps every RPC call
  in `withUserContext(userId)` so `auth.uid()` resolves to the seeded
  owner. Without this the new membership check would have failed all
  existing tests.
- New pg-real test: `rejects with BATCH_UNAUTHORIZED when caller is
  not a member of the company` — outsider user gets explicit refusal.
- New happy-path assertion: `source_type = 'supplier_invoice_paid'`
  on the combined verifikat for supplier batches.

15 unit tests pass on the touched paths. RPC patch applied to remote
via Supabase MCP. Out-of-scope mcp-server changes still parked locally.

Skipped findings (documented in PR comment thread):
  - V8.2.1 ownership pre-check at route layer (RPC enforces it)
  - V4.5 / Art.5(1)(b) narrower API response and event payload —
    typed contracts require the full shapes
  - V2.4 rate-limiting — system-level, applies to all match endpoints
  - A.8.28 client-side RLS reliance — documented architectural choice
  - Direction pre-check at API layer (RPC catches with cleaner code)
  - V16 + Art.32 + Art.5(1)(b) low-severity logging nits

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-29 13:45:41 +02:00
Jakob Wennberg 7bcd46d503 feat(transactions): match overshoot guards + supplier voucher linking (#602) 
* feat(transactions): match overshoot guards + supplier voucher linking

Three changes that together close the "I can't link a bank transaction
to an already-booked verifikat on the supplier side" gap and fix a
latent data-corruption bug on the per-tx match endpoints.

1. fix: clamp paid_amount on match endpoints when tx > remaining

   /api/transactions/[id]/match-{invoice,supplier-invoice} previously
   used transaction.amount wholesale as the paid amount, pushing
   invoice.paid_amount past invoice.total whenever the bank tx was
   larger than what was owed. Both endpoints now reject with
   MATCH_AMOUNT_EXCEEDS_REMAINING / MATCH_SI_AMOUNT_EXCEEDS_REMAINING
   and a structured { transaction_amount, remaining_amount, excess }
   payload that points the user at the future split-payment flow.
   FX branch already clamps to invoice.remaining_amount and is
   unchanged.

2. feat: supplier-side "link existing verifikat" (mirror of #591)

   lib/invoices/supplier-voucher-matching.ts mirrors the customer
   voucher-matching module: finds posted JEs that debit 2440
   (Leverantörsskulder), validates currency + remaining-amount, and
   atomically links them as supplier_invoice_payments rows. New
   /api/supplier-invoices/[id]/{voucher-candidates,link-to-voucher}
   routes wrap it. LinkVoucherPicker gains a mode='supplier_invoice'
   prop so the same component renders both flows. The supplier-invoice
   mark-paid dialog now uses Tabs ("Ny betalning" / "Befintlig
   verifikation") to match the customer-side UX.

3. infra: transaction_voucher_links junction + denorm guard

   Foundation migration for upcoming multi-tx ↔ multi-voucher flows.
   Adds the junction table (with RLS, updated_at, indexes), a
   block_contradictory_invoice_denorm trigger on transactions that
   refuses to set invoice_id/supplier_invoice_id to a value that
   contradicts an existing payment row, and is_transaction_booked(uuid)
   as a single source of truth for "is this tx anchored?" once
   multi-allocation leaves denorm columns NULL. No application code
   uses these yet — they unlock the batch allocation and bulk-book
   flows in follow-up PRs.

Tests: 98 unit tests pass across the touched paths (match-invoice,
match-supplier-invoice, supplier-voucher-matching, link-to-voucher).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(supplier-invoices): PR review — atomic link RPC, computeRemaining edge case, pg-real tests

Addresses the three real issues raised by Greptile on PR #602.

1. (P1) Atomic supplier voucher linking — new
   link_supplier_invoice_to_voucher PL/pgSQL RPC. The TS-side
   linkSupplierInvoiceToVoucher() previously did UPDATE-then-INSERT with
   a manual unconditional rollback. Under concurrent linking against the
   same invoice, request A's rollback could overwrite a sibling B's
   successful write while leaving B's payment row in place. Moving both
   writes into a single PG transaction (one RPC call) lets PG's own
   rollback handle the failure path correctly. TS wrapper now just
   translates the structured RPC return into the lib's Result type.

2. (P1) pg-real tests — tests/pg/transaction_voucher_links.pg.test.ts.
   CLAUDE.md mandates *.pg.test.ts for any PR adding a trigger, RPC, or
   RLS. The Phase 1A foundation migration added all three but had no
   pg-real coverage. Tests now cover:
     - trg_block_contradictory_invoice_denorm refusing contradictory
       UPDATEs on invoice_id and supplier_invoice_id
     - the same trigger PERMITTING a matching UPDATE (no false positives)
     - is_transaction_booked() returning true via journal_entry_id, via
       invoice_payments, and via transaction_voucher_links rows.

3. (P2) computeRemaining edge case — trust remaining_amount whenever
   the column is non-null (including the legitimate 0 for fully-paid
   invoices). The old "> 0" guard fell through to total - paid_amount,
   which under rounding drift could compute a tiny positive residue and
   slip a fully-paid invoice past LINK_SI_VOUCHER_INVOICE_FULLY_PAID.

The fourth Greptile comment (overdue invoices silently get no
candidates) was a misread: 'overdue' IS in the open-state list at
route.ts:35. No code change needed there.

Tests: 100 unit tests pass (16 in the directly-touched paths).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(supplier-invoices): PR review round 2 — broaden AP range, log event failures

Addresses the actionable findings from the compliance-swarm and
Swedish-accounting-compliance bot reviews on PR #602.

1. (swedish-accounting-compliance, high) AP account hardcoded to 2440
   rejected legitimate samlingsverifikationer that debit 2441
   (Leverantörsskulder i utländsk valuta), 2443 (Skuldfakturor), etc.
   BAS 2026 reserves the full 2440–2449 range for Leverantörsskulder.
   The TS-side AP_ACCOUNT constant becomes AP_ACCOUNT_PREFIX ('244')
   used with .like() and .startsWith(). The PL/pgSQL RPC's
   account_number filter becomes LIKE '244%'. The
   LINK_SI_VOUCHER_NO_AP_DEBIT error message updates to reference the
   244x range with examples.

2. (ISO 27001:2022 A.8.15 / OWASP V16) Empty catch on the
   supplier_invoice.paid event emission now logs with log.warn so a
   failure in the downstream reminder/audit subscriber leaves an
   auditable trail without blocking the response.

3. (GDPR Art.5(1)(c)) Documented design rationale for retaining
   select('*') on the post-link invoice re-fetch: the
   supplier_invoice.paid event payload is typed as
   `supplierInvoice: SupplierInvoice` in lib/events/types.ts, narrowing
   would break the subscriber contract. The event stays in-process
   and consumers legitimately need the full context.

Skipped findings:
  - V8.2.1 ownership concerns: route + RPC already filter by
    company_id from withRouteContext; the RPC's WHERE clause covers it.
  - DELETE policy scoping: matches the gnubok pattern across all
    company-scoped tables — any member with write access manages records.
  - transaction_id = NULL on the voucher-link path: by design — the
    flow has no bank tx (the voucher's 1930 line represents it).
  - Reverse-charge VAT (2614/2647) validation on linked vouchers:
    real concern but invasive change; tracked for follow-up.
  - Storno-chain integrity (linking the original of a storno pair):
    edge case; tracked for follow-up.

Tests: 26 unit tests pass in the directly-touched paths. RPC patch
applied to remote via Supabase MCP.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-29 12:31:19 +02:00
Jakob Wennberg ccdfed5fea feat: voucher linking, recovery ops, and salary overrides (#591) 
* feat: voucher linking, recovery ops, and salary overrides

Adds reversible/correction-style write paths that customers and agents have
been asking for, plus per-run salary employee overrides.

Invoice → voucher linking
- POST /api/invoices/[id]/link-to-voucher and
  GET /api/invoices/[id]/voucher-candidates
- lib/invoices/voucher-matching.ts with full + pg test coverage
- LinkVoucherPicker UI in PaymentBookingDialog
- pending_operations.operation_type expanded with link_invoice_voucher
  (medium risk) and a (journal_entry_id, invoice_id) unique guard
- MCP: gnubok_find_voucher_candidates_for_invoice and
  gnubok_link_invoice_to_voucher tools

SIE undo
- POST /api/import/sie/[id]/undo + undo_sie_import RPC
- sie_imports.status gains 'undone'
- ImportResultStep surfaces the action; structured error SIE_UNDO_FAILED

Edit-recreate journal entries
- POST /api/bookkeeping/journal-entries/[id]/edit-recreate
- Bookkeeping detail page wires it into the existing edit flow

Delete-last-voucher clears IB link
- Trigger + pg test ensure deleting the last voucher of a period nulls the
  opening_balance_journal_entry_id link so a re-import lands cleanly

Salary employee overrides
- salary_run_employees gains per-run override fields + migration
- lib/salary/effective-values.ts centralises resolved values; all payslip,
  payment, AGI, KU, and booking routes read through it
- SalaryOverridePanel on the employee detail page

Account classifier
- lib/bookkeeping/account-classifier.ts + tests; AddAccountDialog uses it
- backfill-import-accounts script updated

Misc
- toast: minor styling tweak
- AGI generate-declaration: respect effective values
- structured-errors: new LINK_INVOICE_VOUCHER namespace

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat: add link_invoice_voucher operation type to pending_operations

* feat: refactor salary run calculations and update error handling for SIE imports

* fix: PR review feedback on voucher linking and SIE recovery

pg-real (blocking):
- tests/pg/delete-last-voucher-ib: drop posted_at = now() from the seed
  UPDATE — journal_entries has no posted_at column.
- lib/invoices/__tests__/voucher-matching.pg: seed the posted voucher
  before closing the fiscal period so enforce_period_lock doesn't block
  the INSERT during setup.

voucher-matching error codes and rollback:
- Add LINK_VOUCHER_DB_ERROR (HTTP 500) and return it on real invoice
  UPDATE / payment INSERT failures. Previously these returned
  LINK_VOUCHER_VOUCHER_NOT_FOUND (404) which the pending-op dispatcher
  auto-rejects on transient DB errors.
- Log rollback failures explicitly so an invoice left in a half-linked
  state (advanced status, no payment row) surfaces for manual
  reconciliation instead of disappearing silently.

resyncNextPeriodOpeningBalance ordering:
- Create the new IB first, relink the period FK, then storno the old IB.
  Previously the storno ran first; if createJournalEntry failed the next
  period was left with a reversed IB and nothing to replace it, and
  executeSIEImport swallows the error as a non-fatal warning.

replace_period_opening_balance_link:
- Tighten role check to owner/admin (was owner/admin/member). Matches
  delete_last_voucher and undo_sie_import.

Data minimisation:
- /api/invoices/[id]/voucher-candidates and the matching MCP tools now
  project only the invoice and customer fields the matcher reads, instead
  of returning the full customer row.

Schema bounds:
- SalaryEmployeeOverrideSchema caps each numeric override at 10 MSEK to
  catch typos before they reach the ledger or AGI.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(tests): supply user_id when seeding voucher_sequences

voucher_sequences.user_id is NOT NULL (per the multi-tenant refactor in
20260330130000). The previous test seed only set company_id /
fiscal_period_id / voucher_series, which made the seed fail with a
constraint violation on the latest pg-real run. Pass the same userId
used elsewhere in the seed helper.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(tests): scope delete-last-voucher RPC assertions inside the tx

withUserContext always ROLLBACKs, so any DELETE the RPC performs is
discarded when the callback returns. The previous test then queried
journal_entries via a fresh getPool() connection that only saw the
pre-RPC committed seed state — hence "expected '1' to be '0'".

Move every post-RPC assertion (entry count, period FK clear,
opening_balances_set flip, audit log entry, sie_imports clear) inside
the same withUserContext callback so they observe the uncommitted state
before ROLLBACK fires.

Also fix the sie_imports INSERT: the column is `filename`, not
`file_name`, and `sie_type` is NOT NULL.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(tests): assert against the IB-marker audit row directly

DELETE on journal_entries fires two audit_log writes: the generic
write_audit_log() trigger row ("Deleted journal_entries record") and the
delete_last_voucher RPC's explicit "(was period IB)" entry. Both land
at the same statement_timestamp(), so ORDER BY created_at DESC LIMIT 1
returned the trigger row non-deterministically in CI.

Switch to a presence check with a LIKE filter on the IB marker so the
test verifies what it actually cares about — that the RPC's IB-aware
audit row exists.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(db): set company_id on delete_last_voucher audit_log rows

20260528120000_delete_last_voucher_clears_ib_link.sql inserts directly
into audit_log without setting company_id. audit_log's SELECT policy
filters company_id IN user_company_ids(), so those rows landed with
company_id=NULL and were invisible to every reader — only the generic
write_audit_log() trigger row remained visible. That broke BFL audit-
trail intent: the "(was period IB)" provenance row was never readable.

Republish delete_last_voucher with p_company_id populated on both
audit_log INSERTs (draft path and posted path). Behavior is otherwise
unchanged; the pg-real test for the IB-clear flow now sees the
RPC-written marker row as expected.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Emil <emilmattsson14@gmail.com>
 2026-05-28 21:09:43 +02:00
Jakob Wennberg f53725b20a Agent v1 bundle: TIC v2 onboarding, in-app assistant gating, sidebar nav, MCP fixes (#584) 
* fix(sie-import): accept tab as field separator (Bollbok exports)

The SIE 4 spec allows either space or tab between fields, but
splitSIELine() only treated space (0x20) as a separator. Bollbok
exports tab-separated lines for every record except #RAR, which
silently swallowed all #IB / #UB / #KONTO / #KTYP / #VER / #TRANS
records — imports appeared empty even though the file was well-formed.

Also adds a parser-side diagnostic that emits a warning when raw #IB
or #VER lines are present in the input but parsing produced none. The
previous silent failure is how this bug stayed hidden; the warning
gives the import preview something visible to surface next time.

Verified against two real reproducer files (Sean / Erik Hellqvist):
  erik h 2025.SE (UTF-8): 166 accounts, 66 IB, 4 UB, 11 RES, 95 vouchers, 198 TRANS.
  erik h 2026.SE (CP437): 166 accounts, 66 IB, 4 UB, 0 vouchers.
Both now parse with zero warnings/errors.

Tests:
  + 8 Bollbok-shape tab-separated fixtures (2025 + 2026 quoting variants).
  + 4 silent-failure diagnostic-warning tests.
  All 74 sie-parser tests pass; 155/155 in lib/import; 64/64 downstream callers.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(sie-import): address PR #513 review — strip #KTYP quotes, suppress redundant aggregate warning

Two non-blocking P2 findings from Greptile review on PR #513:

1. #KTYP handler stored fields[2] directly, so Bollbok 2026 exports
   (#KTYP\t1510\t"T") stored '"T"' with literal quotes instead of 'T'.
   Latent defect — accountType is unused downstream today, but my tab-
   separator fix made the quoted-value path reachable. Now routes through
   parseStringField so both Bollbok 2025 (unquoted T) and 2026 (quoted "T")
   land as 'T'.

2. The aggregate "kontrollera fältavskiljare och teckenkodning" warning
   fired alongside per-record 'error'-severity issues for malformed #IB /
   #VER records, producing a misleading hint when the parser had already
   pinpointed the structural problem. Now suppressed when an error-severity
   issue with the same tag already exists.

Test coverage:
  + accountType asserted to be 'T' (not '"T"') in both 2025 + 2026 shapes.
  + VER aggregate-warning test now uses #VER lines without { } blocks
    (silent loss, no per-record error) — the canonical case the diagnostic
    is designed for.
  + New suppression test: bare #VER produces per-record errors AND the
    aggregate warning is absent.

75/75 sie-parser tests pass; 156/156 in lib/import.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* wip: agent chat + composer + memory + document extraction

In-progress work on this branch beyond the SIE-import fixes:
- Specialized accountant agent (composer + intents + chat loop)
- Persistent agent_conversations/messages, agent_profiles, agent_memory
- /chat surface + /onboarding/agent + /settings/agent-memory
- document-extraction extension with status hooks
- MCP server staging refactor + new skills (atoms, bank reconciliation,
  customer onboarding, kreditfaktura)
- pending_operations rejection feedback (category + reason) + realtime
- TIC company profile cached snapshot on companies
- 17 migrations (all additive — see prior conversation analysis)

Parked while branch waits for review/merge. Migrations are already
applied to prod.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* refactor(tic): migrate company-data client from api-core v1 to Lens v2

Swaps the seven TIC company-data endpoints we call from the api-core
paths (`/datasets/companies/{companyId}/...`, `/search/companies`) to
the Lens equivalents (`/companies/{id}/...`, `/search-public/companies`).
Hard cutover; proxy pattern preserved.

Schema shifts handled inside the extension so consumers (TicWorkspace,
Step2CompanyDetails) don't need changes:

- `/companies/{id}/bank-accounts` now returns Bankgirot only — map to
  the existing `{ type, accountNumber, bic }` shape, drop terminated.
- `/companies/{id}/industries` returns a discriminated array — filter
  to `companyIndustryCodeType === 'sni2007'` to preserve v1 behavior.
- `/companies/{id}/phone-numbers` renamed the field to
  `phoneNumberFormatted` (fall back to `e164PhoneNumber`).
- `/companies/{id}/documents` replaces `/financial-report-summaries`;
  filter `type === 'annualReport'` and read nested
  `financialReportMetadata` to rebuild the legacy summary shape.
- `isCeased` is now a top-level boolean; `activityStatus` is an enum.
  Translate enum -> 'ceased' for the workspace's existing check.

BankID identity flow (id.tic.io) is untouched — separate TIC product.

Note: deploy gated on the TIC proxy being flipped to lens-api.tic.io
with an `x-api-key` Lens key.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(tic): expose v2 onboarding & workspace data

Adds six new Lens (v2) fetchers on top of the migration that already
landed in this branch, surfacing the data through /lookup and /profile.

New fetchers in lib/tic-client.ts:
- getFiscalYears          /companies/{id}/fiscal-years
- getAccountingPeriods    /companies/{id}/accounting-periods
- getPayrolls             /companies/{id}/payrolls
- getSignatory            /companies/{id}/signatory
- getRepresentatives      /companies/{id}/representatives
- getCompanyStatus        /companies/{id}/status

/lookup gains a fiscalYear field (current fiscal-year configuration)
so onboarding Step 2 can skip manual MM-DD entry. CompanyLookupResult
extended with optional fiscalYear; consumers without it keep working.

/profile gains five new sections on TICCompanyProfile:
- fiscalYear + fiscalYearHistory   current + deduped period list
- signatory                        firmateckning descriptions
- board + representatives          board-composition summary + active
                                   officers (positionEnd in future)
- payrolls                         payroll2 array newest-first, with
                                   deviation vs annual-report
- statuses                         current+historical status entries
                                   with red/yellow/green/neutral color

TicWorkspace renders the new data as four cards (Status, Fiscal year +
Signatory, Board + Representatives, Payroll history) plus a Badge
mapping for the traffic-light status color.

Tests: 52 -> 60 passing. Added unit tests for the new fetchers' v2
paths, fiscal-year auto-fill in /lookup, and full v2 profile coverage.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(onboarding,agent): lean on TIC v2 to skip Steps 1 & 3 and sharpen Opus

Three small wins that unlock more of the v2 cutover. No new endpoints — the
data was already in the snapshot, just not flowing where it should.

Step 1 (entity_type) — deep-link path only:
- /lookup now returns `legalEntityType` and `registrationDate` (added to
  CompanyLookupResult).
- /onboarding/page.tsx does a server-side /lookup prefetch when
  ?org_number= is present (BankID picker path), maps "AB"/"EF" to the
  EntityType enum, and seeds Step 1's radio. Falls through silently for
  unsupported codes (HB, KB, …) and on TIC errors.
- WelcomeOnboarding hydrates ticLookup state from the server prefetch so
  Step 2's debounced client fetch and Step 3's first-year inference both
  have data on first render — no flash.

Step 3 (is_first_fiscal_year) — every path:
- deriveFirstYearDefaults() parses ticLookup.registrationDate and returns
  { isFirstFiscalYear, firstYearStart } when registered <12 months ago.
  Step 3's initialData picks it up; the user only confirms the end date.
- Settings value wins when present so existing users with a saved choice
  don't get overridden.

Composer prompt:
- redactTic allowlist was the bottleneck — it stripped beneficialOwners,
  signatory, board, representatives, payrolls, statuses, fiscalYear
  before Opus ever saw the JSON. Existing filterRedundantQuestions
  ownership logic was effectively dead because the data path was severed.
  Expanded allowlist to include those v2 sections; kept bankAccounts/
  email/phone/fiscalYearHistory/financialReports out (token cost > signal).
- SYSTEM_PROMPT now documents each v2 section and the rules Opus should
  apply: payroll signal switches from "registration.payroll" to "actual
  payrolls[] filings" (kills the false-positive swedish-payroll selection
  for newly registered employers); beneficialOwners[] becomes the
  authoritative ownership source (single owner → FMB modifier; multiple →
  multi-owner); statuses[] isCeased/red triggers an uncertainty_note.

Tests: 4112 unchanged. Build: green. No schema or migration changes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(agent): onboarding polish + composer signal fixes from first-run feedback

UX:
- AgentOnboarding: drop the 10s "Hoppa över — fortsätt med standardval"
  escape hatch. The fallback path runs automatically on timeout; the
  manual skip just teased users into a degraded build.
- ReviewCard step 2 title: "Stämma av detaljerna" → "Stäm av detaljerna"
  (imperative form matches the rest of the steps).
- Drop em-dashes from user-visible Swedish strings in AgentOnboarding +
  ReviewCard (fallback labels, subtitles, placeholder, error message,
  final CTA). Em-dashes survive in code comments only.
- "Fråga min revisor" → "Fråga min assistent" everywhere it surfaced:
  AgentTrigger, AgentSparkleButton, ReviewCard preview, ReviewCard
  fallback comment, general.help intent buttonLabel + prompt text.
- AgentTrigger / AgentSparkleButton / EmptyState.AgentHelpLink /
  TransactionInboxCard ask-button all gated on identity.isVerified.
  Pre-onboarding users no longer see the floating FAB or per-page
  Sparkle buttons. AgentSheetProvider.identity gained an isVerified
  field; (dashboard)/layout.tsx selects agent_profiles.verified_at and
  passes it through.

TIC verksamhetsbeskrivning:
- tic/index.ts /profile: /companies/{id}/purposes returns every
  historical verksamhetsföremål filing. Picking [0] was returning the
  oldest "äga och förvalta" holding-company boilerplate for companies
  whose later filings narrowed the purpose ("tillhandahålla
  företagskrediter och finansiella teknologilösningar"). Sort the
  array by lastUpdatedAtUtc desc and take the most recent non-empty
  purpose.

Composer banking signal:
- loadBankingSummary now reads journal_entry_id alongside
  description/amount/date and returns per-counterparty `direction`
  ('in' | 'out' | 'mixed') and `has_unbooked` (any row not yet booked).
  Aggregate `unbooked_count` accompanies the rollup.
- buildUserPrompt emits each counterparty as
  `Name: 12 345 kr (ut, OBOKFÖRD)` so Opus can tell income from cost
  on sight and tell which counterparties are still open questions.
- SYSTEM_PROMPT now explicitly forbids verification questions about
  counterparties whose direction is unambiguous AND status is 'bokförd'.
  Should kill the regressions from the first agent build:
  * "Konsult, J 98 565 kr — intäkt eller kostnad?" when the amount is
    clearly negative.
  * "ALMI AB 493 000 kr — lån eller bidrag?" when the transaction is
    already categorized.

Tests: 4112 unchanged. Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(agent,ui): representation needs deltagare+syfte, drop duplicate doc icon

Representation booking:
- transaction-categorization prompt now requires the agent to capture
  participants (name + company) AND purpose before staging a
  representation categorization. SKV's representationsregler + ML 8 kap
  require the verifikation to document who attended and what the
  meeting was about; without that the avdrag is denied and the post
  should be booked as non-deductible / personalkostnad.
- The agent confirms back in plain text (audit trail in the chat),
  writes the deltagare + syfte to gnubok_remember_fact (long-term),
  THEN stages. Saknas deltagare/syfte: explicitly tell the user the
  avdrag won't go through and offer the non-deductible alternative.
- Known gap (followup, not this commit): the staged op's journal entry
  description doesn't yet carry the deltagare text. Until we add a
  `notes` field to gnubok_categorize_transaction, the audit trail
  lives in chat + agent_memory only.

TransactionInboxCard duplicate attachment indicator:
- Drop the FileCheck2 "open document" button from the trailing slot.
  TransactionAttachmentIndicator (Paperclip) next to the description
  already opens the underlag on click. Two icons doing the same thing
  was noise. Cleaned up the unused state (isOpeningDoc, hasAttachment,
  handleOpenAttachment) and dropped now-unused imports (FileCheck2,
  useToast).

Tests: 4112. Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(agent,nav): notes on verifikation + redesigned sidebar

Audit-trail notes for representation:
- gnubok_categorize_transaction gains an optional `notes` string.
  Threaded through stagePendingOperation → commitCategorizeTransaction →
  createTransactionJournalEntry, which now appends notes to the entry's
  description (capped at 500 chars). The verifikation an external auditor
  reads now carries deltagare + syfte directly — not just chat history /
  agent_memory.
- transaction-categorization prompt updated: representation flow now
  REQUIRES the agent to pass deltagare+syfte via the notes parameter.
  Without it the booking is non-deductible / personalkostnad per SKV.

DashboardNav redesign:
- Top section: flat, no header — Hem (/chat), Underlag (was
  Dokumentinkorg), Transaktioner, Granskning. Always visible; the inline
  badge on /pending shows the count when there are pending ops.
- Mid section: four collapsible dropdowns (Försäljning, Inköp,
  Redovisning, Personal). Each auto-expands when the active route lives
  inside it. KPI moved from main to Redovisning. Extension nav items
  (TIC workspace, etc.) fold into Redovisning.
- Bottom-left: new account popover (DropdownMenu, opens upward) holding
  CompanySwitcher, Inställningar, Hjälp, Support, Logga ut. Replaces
  the old top company-switcher card + the bottom Support/Logout block.
- Mobile drawer mirrors the new structure: top items as flat list,
  same four dropdown groups, separate "Tillägg" section when
  extensions exist, "Mitt konto" section at the bottom.
- i18n: invoice_inbox label renamed "Dokumentinkorg" → "Underlag"
  ("Documents" in en). New keys: mitt_konto, group_extensions.

Tests: 4112. Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(nav): unhide Leverantörer under Inköp

The /suppliers entry existed in navItems but was marked hidden — leftover
from when the supplier list lived elsewhere in the IA. Removing the
hidden flag puts Leverantörer in the Inköp dropdown alongside
Leverantörsfakturor.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(nav): CompanySwitcher back to top-left, user account moves bottom-left

The previous pass collapsed both concepts into the bottom popover. They
mean different things: the company is the org context everything below
operates against (top-of-sidebar, scannable); the user is the
account-holder (bottom-of-sidebar, where settings/logout live).

- (dashboard)/layout.tsx: fetch profiles.full_name alongside the
  existing identity queries; pass userName + userEmail into
  DashboardNav.
- DashboardNav: restore CompanySwitcher at the top of the sidebar
  (pre-redesign placement). Bottom-left popover trigger now shows the
  signed-in user's name + single-letter initial (accountInitial helper
  falls back to email's first char, then "?"). Popover header carries
  full name + email; items unchanged (Inställningar, Hjälp, Support,
  Logga ut). CompanySwitcher removed from inside the popover — nested
  dropdowns were awkward and the top placement is where it belongs.

Tests: 4112. Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(pending): trim the agent context strip

The row-level AgentContextStrip on /pending was rendering the model
name (eu.anthropic.claude-sonnet-4-6) and the full atoms array
(horizontal/swedish-vat, vertical/konsult-it, …) inline, which made
each row 60–80 chars of mostly-the-same metadata. Reviewers never
scan that text; they scan amounts and decide approve/reject.

Now the strip shows only the conversation deep-link
(Konversation #<short id>) — the one piece that's actually useful for
diving into context. Model + atoms remain available in agent_metadata
for debugging surfaces; they're just not in the list view.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(agent): shared ground rules + paragraph breaks after tool calls

Two regressions surfaced in real usage. Both are systemic.

Shared agent ground rules:
- /chat surface (general.help) was happily inventing four-digit BAS
  account numbers ("Debet 6212 - Molntjänster…", "Kredit 2614 - Ingående
  moms…") and proposing booking decisions on invoices it had never
  seen, with no follow-up questions about currency/scope/etc.
- transaction-categorization had those rules baked into its prompt;
  general-help / bokslut-step / invoice-draft / supplier-invoice-review
  / verifikation-draft / vat-review never inherited them.
- Extracted lib/agent/intents/shared-rules.ts with five cross-cutting
  rules: underlag first (check inbox + ask user to upload to
  Dokumentinkorgen when missing), ask follow-ups when ambiguous, never
  write four-digit BAS account numbers in chat (category names only),
  cite atoms / load skills (don't guess), check counterparty history
  before proposing.
- Injected renderAgentGroundRules() into all six intents above.
  transaction-categorization left alone — it has more detailed inline
  rules tied to its specific underlag-flow.

Paragraph break after tool calls:
- text_delta from the model often resumes after a tool call without a
  leading newline ("kategoriseras." → gnubok_query_journal runs → "Inget
  historik hittades…" appended directly). Markdown rendered the
  concatenation as one paragraph.
- AgentChat text_delta handler now inserts \n\n when (a) the buffer
  ends with text content, (b) the incoming delta starts with text
  content, (c) at least one tool call has run, and (d) the buffer
  doesn't already end with a blank line.

Tests: 4112. Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(nav): default-open dropdown groups; closing is per-user

Dropdowns started collapsed which meant first-time users had to open
each group to discover what's inside. Inverted the state: default open,
user can collapse, active route still forces a group open.

- manualExpanded → manualCollapsed (semantics flip)
- toggleGroup unchanged externally; flips the bit
- isGroupExpanded returns !manualCollapsed[g] || hasActiveChild

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(agent): rate-safe v1→v2 TIC upgrade, counterparty defaults, profile settings

Three pre-ship quality wins.

Rate-limit-safe TIC v2 upgrade:
- The /profile endpoint fans out to ~13 Lens calls; the account has a
  ~3000/mo ceiling. Force-refreshing every pre-v2 (v1) snapshot across
  the customer base would blow the budget.
- ensureTicSnapshot gains an `upgradeV1` flag. A cached snapshot still
  inside the 7-day window is re-fetched only when (a) the caller passes
  upgradeV1 AND (b) the snapshot is v1-shaped (missing the v2-only
  `statuses` key). Gated to the two agent-onboarding call sites — a
  deliberate, once-per-company action and the only consumer of the v2
  sections. Workspace + signup keep the natural 7-day staleness, so the
  v1→v2 migration is lazy and bounded to companies actually building an
  agent.

Known-counterparty defaults (shared-rules):
- Agent now proposes a sensible default for well-known counterparties
  instead of asking the same question monthly: Almi → lån, Tillväxtverket/
  Vinnova/EU-stöd → bidrag, Skatteverket → skatt/avgift or återbäring,
  Bolagsverket → avgift, Försäkringskassan → ersättning, EF private
  withdrawal → eget uttag. Stated as an assumption the user can correct,
  not a hard rule — underlag/history still wins.

Företagsprofil settings page:
- New /settings/agent-profile (Företagsprofil / "Company profile"):
  view + edit the agent's company profile after onboarding — assistant
  name + avatar, the profile summary the agent reasons from, and a
  read-only chip view of loaded specialities (atoms). Backed by the
  existing GET/PATCH /api/agent/profile.
- New GET /api/agent/atom-titles?ids= resolves atom slugs → human titles
  for the chips (registry is globally-readable reference data).
- Added to SettingsSidebar; i18n keys agent_profile (sv "Företagsprofil"
  / en "Company profile").

Note: /chat already redirects unverified users to / (chat layout guard),
and / renders WelcomeGate → /onboarding/agent. No redirect work needed.
AgentSetupBanner.tsx is orphaned dead code (WelcomeGate superseded it).

Tests: 4112. Build: green. Both new routes compile.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(nav,agent): Hem=Översikt + separate Assistent button; memory dedup

Nav restructure:
- "Hem" now points to / (Översikt dashboard) again, not /chat. The agent
  chat gets its own top-level nav entry "Assistent" (Sparkles icon) → /chat.
  Mobile bottom nav mirrors this (Hem / Assistent / Transaktioner).
- / restored to render DashboardContent (the Översikt) for built-agent
  users instead of redirecting to /chat. Users who haven't built their
  assistant yet still get WelcomeGate (the build-agent checklist); once
  verified, / shows the dashboard. Chat is reachable anytime via its nav
  entry. Restored main's dashboard data-fetch; added an agent_profiles
  verified_at probe to drive the WelcomeGate branch.
- i18n: nav.assistant ("Assistent" / "Assistant").

agent_memory dedup (gnubok_remember_fact):
- The agent re-remembers the same fact constantly (e.g. "Vercel = omvänd
  skattskyldighet" on every Vercel categorization), which would bloat
  agent_memory with paraphrases over months.
- Before insert, compare the incoming fact against the 300 most-recent
  active memories by word-set Jaccard similarity (lowercased, punctuation-
  stripped, stopwords dropped). A near-duplicate (≥0.82) is treated as
  already-known: bump its relevance toward the new score + refresh
  updated_at instead of writing a new row. Embedding-free, zero added
  latency beyond one bounded SELECT.

Tests: 4112. Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(agent,nav): företagsprofil=Bolagsuppgifter, avatar nav icon, dedupe greeting

Företagsprofil settings page (the right content this time):
- Replaced the agent atoms/summary panel with CompanyProfileView — a
  read-only "Bolagsuppgifter" view of the cached TIC company snapshot
  (name, org-nr, form, address, F-skatt/Moms/Arbetsgivare, SNI, bank,
  verksamhet, employees, latest financials, status traffic-lights,
  fiscal year, firmateckning, företrädare). Server component reads the
  companies.tic_snapshot column directly — no extension import, stays
  inside the core-build boundary.
- Route renamed /settings/agent-profile → /settings/company-profile.
  Removed the old AgentProfilePanel + the now-unused /api/agent/atom-titles
  endpoint.

"Assistent" nav icon = the agent's chosen avatar:
- DashboardNav reads agent identity from AgentSheetProvider and renders
  the onboarding-chosen avatar for the /chat ("Assistent") entry across
  desktop sidebar, mobile drawer, and mobile bottom nav. Falls back to
  the Sparkles glyph pre-onboarding (no avatar yet).

Nav cleanup:
- Dropped the beta badge from Underlag.
- Filtered the TIC workspace (/e/general/tic, "Företagsprofil") out of
  the nav — the same Bolagsuppgifter now lives under Inställningar →
  Företagsprofil, so it shouldn't appear in two places.

Doubled intake greeting fix:
- /chat/intake fires an invoke with no conversation_id, then swaps the
  URL to /chat/[id] the instant the `conversation` event lands — which
  can beat the greeting being persisted. /chat/[id] then hydrated with 0
  messages and, because the auto-fire guard keyed on (id && messages>0),
  fired a SECOND invoke on the same conversation → two greetings.
  Guard now keys on conversation-id presence alone: a set id means
  resume, never bootstrap. Closes the race.

Tests: 4112. Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(agent): paragraph-break-after-tool split words mid-stream

The earlier "insert \n\n when text resumes after a tool call" heuristic
re-evaluated on EVERY text_delta (any delta not starting/ending with
whitespace, once a tool had run). Streaming deltas arrive in sub-word
chunks, so it injected breaks between fragments of the same word:
"minnes\n\nno\n\nterna", "kund\n\nrep\n\nresentation".

Replace the per-delta heuristic with a consume-once ref:
- tool_use sets breakBeforeNextTextRef = true
- the next text_delta consumes it: prepends \n\n exactly once (only when
  the buffer has content, doesn't already end in whitespace, and the
  delta doesn't start with whitespace), then clears the flag

So the break fires once per tool→text resume, never mid-word.

Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(agent): much shorter replies, representation headcount + VAT cap, dot separator

Brevity (system-prompt Svarsformat — affects every reply):
- Hard "korthet är regel nummer ett": aim for 2-4 sentences, lead with
  the answer/action, no warm-up ("Här är vad som gäller…"), don't derive
  VAT in prose, don't restate what the approval card shows, one question
  at a time. The agent was writing textbook-length essays.

Representation rule now in shared-rules (so verifikation-draft, vat-review,
etc. all get it — previously only transaction-categorization had it, which
is why the verifikation flow guessed 25% VAT and skipped the cap):
- Require ANTAL deltagare (headcount), not just one name — the moms
  deduction is per person (underlag cap 300 kr/person ex moms).
- Use the receipt's ACTUAL VAT rate (usually 12% on food), never assume
  25%.
- Meal representation isn't income-tax deductible (post-2017); whole cost
  booked as non-deductible representation.

Verifikation description separator:
- createTransactionJournalEntry appended notes with an em-dash
  ("Utlägg Eatnam — Deltagare:…"), violating house style. Switched to a
  middle dot " · ". journal_entries has no separate notes column — the
  description IS the BFL verifikationstext / audit field, so deltagare +
  syfte correctly live there.

Tests: 4112. Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(settings): tidy Bolagsuppgifter — no status colours, clean firmateckning

From first-look feedback on the Företagsprofil page:

- Status: dropped the coloured traffic-light badges (red/yellow/green).
  Per the design system semantic colour is data-only, never chrome, so
  status now renders as plain label + date. Also filtered to dated
  entries only — Bolagsverket emits flags like "Har aldrig varit verksam"
  with no date that read as noise next to the real status. Ceased status
  gets muted destructive text (the one chrome colour the system keeps).

- Firmateckning: the source text carries ">" list markers and crams
  several rules onto one line, and repeats "Firman tecknas av styrelsen"
  across rows. cleanSignatory() strips the markers, normalises whitespace,
  splits run-on "Firman tecknas …" clauses onto separate lines, and the
  render dedupes — so each rule reads as its own sentence.

Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): inbox items expose all terminal links + processed flag

The Eatnam receipt was booked against its bank transaction (so the inbox
row had matched_transaction_id + created_journal_entry_id set), yet the
agent reported it as loose/unmatched and a duplicate risk. Root cause:
gnubok_list_inbox_items only selected and returned matched_supplier_id +
created_supplier_invoice_id — the supplier-invoice path. The
transaction-match and direct-journal-entry paths were invisible, so any
receipt cleared via /transactions looked unprocessed.

- list_inbox_items now selects + returns matched_transaction_id and
  created_journal_entry_id alongside the supplier fields, plus a derived
  `processed` boolean (true when ANY of the three terminal links is set).
- New unprocessed_only=true input filters to items with no terminal link
  — the "what still needs handling" view that prevents the agent from
  flagging already-booked docs as duplicates. (Fetches a wider window
  then filters client-side so limit applies post-filter.)
- Description updated to document the processed semantics, within the
  280-char tool-description budget.

The DB linkage itself already worked: /transactions attach-document sets
matched_transaction_id, and commitCategorizeTransaction stamps
created_journal_entry_id. This was purely a read/surface gap.

Tests: 4112 (+ MCP description guard). Build: green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): repair stage-but-never-commit tools + consolidate tool surface

- post_annual_depreciation AND reverse_entry were never in the pending_operations operation_type CHECK, so both staged then died with check_violation at INSERT. Add the CHECK migration, a commitPostAnnualDepreciation executor (reusing commitAnnualPostings), risk tier, and the PendingOperationType union member.
- Salary tools de-risked: calculate_salary_run calls runSalaryCalculation() directly (no self-fetch/forged cookie); create_salary_run uses a transactional create-run helper with compensating delete; generate_agi actually generates + persists the declaration.
- import_sie parses + validates at stage time with a content-rich preview (company, fiscal year, voucher/account counts, balance) instead of a blind byte count.
- batch-match-invoices passed user.id where companyId was expected (silently matched zero).
- VAT report+widget merged behind render_ui; gnubok_search_tools ranks by relevance; gnubok_feedback readOnlyHint corrected; tools/list instruction text fixed; income decision-tree + GL/query_journal cross-refs added.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(agent): load skill atom bodies from the DB so they survive the build

Skill bodies were read from disk at runtime (.claude/skills/**/SKILL.md); on Vercel the dynamic readFile path isn't traced into the lambda and on Docker .claude/ is excluded, so atoms loaded EMPTY in production — a despecialized agent. Inline the bodies into agent_atom_registry instead:
- Migration adds body + mcp_exposed columns; a build-time generator (scripts/generate-skill-bodies.ts) emits a deterministic dollar-quoted seed migration with a content-hash manifest + --check CI guard.
- Read sites (mcp-server atoms.ts, chat system-prompt.ts, composer prewarm) read body from the DB, with a dev-only disk fallback. mcp_exposed curates which atoms the MCP exposes (swarm-* never become atoms).
- The seed script + generator share scripts/lib/atom-discovery.ts; estimated_tokens now reflects SKILL.md only.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(agent): safe the in-app assistant — gating, FAB de-confliction, rate limit, friendly errors

- Hide all agent entry points until verified_at: the Assistent nav tab (sidebar + mobile) and the agent-memory settings tab now match the floating FAB's gate.
- FAB de-confliction: /kpi -> kpi.explain and /bookkeeping/year-end -> bokslut.step so the floating button opens the SAME assistant as the page button (no two-agents-on-one-page).
- Generous per-user rate limit (30/min, 1000/day) on /api/agent/invoke, /onboarding/stream, /composer via a new agent_rate_counters table + check_and_increment_agent_quota RPC; fails open. Bounds runaway Bedrock spend without touching normal users.
- Friendly errors: Bedrock 429/timeout/5xx normalized to Swedish (friendlyModelError) in run-turn + the invoke route; the chat client surfaces the server's friendly message instead of a raw HTTP status.
- /chat/new validates ?intent= against the registry so bad deep-links fall back to general.help instead of rendering a broken-looking error.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(agent): keep /chat read-only — redirect categorization + swap the "categorize" suggestion for a VAT-report question

general.help (the /chat assistant) is read-only, but it still gave per-transaction bokföringsförslag in prose and asked "godkänner du dessa?" — an analysis the user can't act on (no write tool, no per-tx underlag). Strengthen the prompt to redirect categorization/bokföring to the per-transaction flow (open the transaction -> "Fråga om denna transaktion", where the agent sees the underlag and stages a real ApprovalCard); a short overview is still allowed. Add a guard test locking in no-write-tools + the redirect language. Swap the /chat empty-state "Hjälp mig kategorisera" chip (which lured users into exactly this dead-end) for a VAT-report question the read-only assistant can actually answer.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* refactor(pending): declutter the review queue rows + header

Fold the conversation deep-link onto the actor label (drop the separate
"Konversation #xxxx" strip and its icon), hide the quick-pick when there's
only one operation type (it duplicated "Markera alla"), and drop the "(0)"
from the disabled bulk-approve button.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(vat): enhance VAT handling by integrating document validation and improving error messaging

* feat(settings): add assistant knowledge surface + consolidate settings tabs

Expose the agent's skill atoms (agent_atom_registry) in a read-only surface
beside the existing memory view, and tighten the settings tab bar from 14 to
10 tabs.

- New GET /api/agent/skills + AgentSkillsPanel: lists active, mcp_exposed
  atoms grouped by tier (Kärnkompetens / bransch / bolagssituation), flags
  which are active for the company from agent_profiles, and lazy-loads each
  SKILL.md body on expand.
- New /settings/assistant tab with a Minne/Kompetens toggle (?view=skills);
  /settings/agent-memory and /settings/agent-skills redirect into it.
- Merge Företagsprofil (TIC snapshot) into the Företag tab via
  CompanyProfileSection; /settings/company-profile redirects.
- Merge Skatteverket-anslutningen into the Skatt tab — OAuth returnTo and the
  callback toast now target /settings/tax; /settings/skatteverket redirects.
- Drop the Säkerhetsbackup tab (already under Importera/Exportera).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(inbox): keep booked underlag out of the unmatched queue + widen match window

- categorize: after booking an inbox underlag onto a verifikat, backfill the
  inbox row's matched_transaction_id + created_journal_entry_id so it stops
  showing as unmatched (mirrors the /attach-document paperclip path).
- TransactionMatchPicker: bias the candidate window forward (60d before →
  180d after the invoice date) so late payments aren't dropped before scoring,
  and widen the ranking date tolerance to 120d so the true match floats to the
  top instead of collapsing to "Svag match". Fix "okatigoriserade" typo.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* wip: bundle in-progress branch work + agent onboarding chat optimizations

Captures the uncommitted work-in-progress on this branch so it lives on the
remote. Heterogeneous changeset — bundled as one commit since the work was
already entangled across files.

Headline change in this commit (from this session):
- Remove the double interview in agent onboarding. Phase B's verification-
  question form stepper is gone — the Phase C chat (onboarding.intake) now
  owns the entire interview and reads the composer's verification_questions
  server-side as its question bank.
- ReviewCard collapses from 3 steps to 2 (meet → review-and-confirm) with
  value-first ordering: profile + "vad jag kan hjälpa dig med" + facts +
  optional seed note. CTA reads "Möt {namn}" to signal the chat follows.
- ChatIntakeStarter handoff subcopy updated to match reality (assistant
  greets first; user can leave anytime).
- Stamp agent_profiles.intake_completed_at server-side in
  app/api/agent/invoke/route.ts on the first user-typed reply in any
  onboarding.intake conversation (idempotent IS NULL guard, best-effort).
  Closes the previously dead-write column and unlocks the opportunistic-
  follow-up hook the migration anticipated.

Plus in-progress branch work being carried forward (not introduced here):
agent runtime + intent prompts, composer + atom-discovery scripts, MCP
server skills surface, onboarding flow components, dashboard/inbox tweaks,
two new agent_atom_registry migrations, additional agent-chat tests.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* refactor(agent): drop inline "Fråga assistenten" affordances — rely on the FAB

The bottom-right "Fråga {namn}" FAB (AgentTrigger) is already route-aware
and picks the right intent per page, so duplicating it as inline page-
header buttons and empty-state links is noise. Removed:

- EmptyState `agentHelp` link ("Eller fråga {namn} hur du kommer igång")
  + the AgentHelpLink component + agent_default_name/agent_ask_link i18n
  keys + the agentHelp props on EmptyInvoices/EmptyCustomers/EmptyTransactions.
- AgentSparkleButton on /bookkeeping (verifikation.draft) and /kpi
  (kpi.explain) page headers.

The FAB stays — when verified, it appears on those routes and routes to
the right intent automatically.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(agent): gate the last two ungated "Fråga assistenten" affordances

Both surfaces previously called useAgentSheet directly without checking
identity.isVerified, so they appeared pre-onboarding (everywhere else the
FAB / sparkle buttons / /chat / Assistent nav are all gated on verified_at).

- Settings page header: remove the "Fråga {namn}" pill entirely. The FAB
  covers /settings routes route-aware (settings.help) — no need for a
  duplicate inline trigger.
- Invoice inbox transaction picker: hide the "Fråga assistenten" button
  when the agent isn't built. Done at the parent (InvoiceInboxWorkspace)
  by passing onAskAssistant only when identity.isVerified is true; the
  child renders the button only when the callback is present.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(tic,onboarding,agent): single-call TIC lookup + director-aware narrative voice

- TIC: collapse the company lookup from 6 endpoint calls to 1
  (search-public already exposes sniCodes, bank accounts, emails, phones,
  and registration flags). Derive fiscal-year MM-DD from
  mostRecentFinancialSummary; newly-registered companies fall through to
  the client's first-year defaults.
- Onboarding: BankID picker no longer auto-provisions companies. Every
  pick routes through the wizard with orgnr (and entity_type via the
  CompanyRoles match) prefilled; F-skatt/VAT/address get confirmed in
  steps 2-4 instead of being auto-fetched. createCompanyFromOnboarding
  reuses CompanyLookupResult and adds a defensive top-level catch so
  server-action errors surface to the UI instead of being redacted.
- Agent composer: loadUserDirectorship() checks BankID CompanyRoles for
  a director-like position (ceo/boardMember/chairman/externalSignatory,
  active) before the narrative uses second-person ownership voice
  ("Du driver…"); unknown users get neutral third-person voice so we
  never put ownership words in the user's mouth.

Tests cover loadUserDirectorship, narrative voice, tic-fetch path,
onboarding page, and updated TIC client + lookup/profile suites.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(tic): extend agent-onboarding TIC budget to 10s + backfill stranded org_numbers

The 5s TIC fetch timeout aborted client-side before the upstream Lens
fan-out (~13 calls) could complete, but the in-flight upstream calls
still counted against quota — actions.ts already documents ~530 wasted
calls from this in May. Same bug still applied to the agent-onboarding
stream path. Adds an optional `timeoutMs` to `ensureTicSnapshot` so
deliberate wait-screen callers (agent onboarding stream) can run with
10s while background/dev callers stay on the conservative 5s default.

Page-level server fetch (page.tsx) intentionally stays at 5s to avoid
blocking TTFB without a visible progress affordance.

Backfill migration mirrors `company_settings.org_number` to
`companies.org_number` for the 105 cases where it's safe (after dedup
+ conflict filtering). 56 of those are on active companies — unblocks
duplicate guards, SIE/SRU exports, and TIC fallback chain. Zero TIC
API calls — pure data move. Idempotent.

Also sweeps a pre-existing SSRF guard on the stream route's origin
derivation that was sitting unstaged in the working tree — it lives in
the same diff hunks as the TIC budget change and couldn't be split cleanly.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* wip: bundle in-progress branch work

Sweep up uncommitted agent/MCP/RLS work-in-progress so the branch is fully
backed up to origin. Not reviewed in detail — committed as-is to preserve
working state alongside the TIC fixes in the previous commit.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(agent): tag the "Bygg din bokföringsassistent" CTA as Beta

Adds a Beta badge next to the assistant-setup heading on the dashboard
banner, dashboard inline card, and onboarding checklist row. Also drops
the stale "Gratis i 30 dagar" subline from the dashboard card.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(build,migrations): PendingOperationType salary ops + resolve migration version collisions

PR #584 went red on three things:

1. core-only build / Vercel: `lib/pending-operations/commit.ts:2666` switched on
   'create_salary_run' and 'generate_agi' but `PendingOperationType` was missing
   both literals. Add them to the union.

2. Supabase preview: migration version 20260526120000 collided with main's
   newly-merged 20260526120000_fix_replace_sie_import_hard_delete.sql.
   Bump the branch's pair to 20260526120050 / 20260526120051 — still ahead of
   20260526120100_restvardeavskrivning so ordering is preserved.

3. 20260527170000 was used twice on this branch
   (_agent_rls_with_check + _journal_entry_no_doc_required). Bump the second
   to 20260527170100 so the pair stays orderable and Supabase doesn't choke
   on the duplicate schema_migrations PK.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(ci): reword comment so core-only guard stops flagging it

The "Check no core imports from extensions" step greps for the literal
\`from '@/extensions/\` across lib/, app/api/, components/. A comment in
lib/agent/composer/tic-fetch.ts quoted the exact pattern verbatim to
explain *why* the file does a self-fetch instead of importing the TIC
extension directly — which the grep matched even though no actual
import exists.

Rewrite the line to keep the same meaning without the literal pattern.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Emil <emilmattsson14@gmail.com>
 2026-05-28 11:27:01 +02:00
Mattsson a9b43ebeb7 Bug/vat selection warning (#583) 
* refactor: update VAT handling logic for non-registered sellers and improve related comments

* chore: gate automated email flows behind 503 responses

Disables user-facing access to invoice payment reminders and salary
payslip email sending. Underlying lib code (reminder-processor,
PDF templates, notification_settings) is preserved for easy re-enable.

- Invoice reminders cron route returns 503; settings UI section removed.
- Payslip send route returns 503; original implementation kept as
  _sendPayslipsImpl for future re-enable.
- Push notifications were already extension-disabled, no change needed.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* chore: remove Recapt feedback widget

Strips the third-party Recapt SDK and its floating feedback bubble from
the app. The in-app contact form keeps working via the existing email
channel (/api/support/contact). Drops the Recapt entries from the CSP
and the subprocessor list in the privacy policy.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat: reject meaningless rättelser in correctEntry

Guard against zero-economic-effect corrections in the storno engine:
- Reject when proposed lines net to zero on every account (e.g. 1930
  debit 100 / 1930 credit 100), which would erase the original posting
  without representing any affärshändelse (BFL 5 kap. 5 §).
- Reject when proposed lines are an exact multiset match of the original
  entry — a rättelse must actually change something.

New MeaninglessCorrectionError wired through bookkeepingErrorResponse
(HTTP 400) and the Swedish error translator.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat: add date-range picker to resultat- and balansrapport

Adds optional from/to date filtering to the four operational financial
reports (resultatrapport, balansrapport, income-statement, balance-sheet)
so users can view a month, quarter, or custom range inside a fiscal year
without leaving the report. Defaults to YTD; "Hela året" preserves the
prior full-period behaviour (URL-identical, cache-stable).

- trial-balance engine accepts optional fromDate/toDate, rolling prior
  in-period activity into IB and clamping period activity to the window
- 12 API routes accept and validate from_date/to_date query params
- ReportDateRange chip picker persists preset per company, only renders
  on the four relevant tabs
- FiscalYearSelector now emits the period object so the range picker
  has bounds without an extra fetch
- PDF/XLSX filenames reflect the chosen range
- Resultatrapport drops the prior-year column when narrowed (full-year
  vs partial-year would mislead)
- 11 new tests (engine + parser); all existing report tests pass

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat: add support for marking journal entries as "no document required"

- Introduced a new sidecar table `journal_entry_no_doc_required` to track entries that do not require separate documentation (e.g., bank fees, interest).
- Implemented API routes for creating and deleting exemptions, including validation and authorization checks.
- Added a toggle component in the UI to allow users to mark entries as exempt, with an optional reason.
- Updated relevant tests to cover the new functionality, including RLS checks and cascading deletes.
- Enhanced existing schemas and types to accommodate the new `vat_amount` field for supplier invoice items.

* fix: address PR review findings on no-doc-required + VAT changes

- pg-real cascade test wraps DELETE in gnubok.allow_delete='true' txn so the
  immutability trigger bypass fires (mirrors delete_last_voucher RPC).
- Clamp supplier-invoice item vat_amount to <= line_total * vat_rate via Zod
  refinement (with 1-öre rounding tolerance) so the manual override can't
  inflate the 2641 debit beyond the statutory ceiling.
- groupVatByRate falls back to line_total * rate when stored vat_amount is 0
  with a positive rate, so legacy/import paths leaving the column at its
  NOT NULL DEFAULT 0 don't silently understate ruta 48.
- ReportDateRange todayIso() and preset endpoints use local date components
  instead of toISOString() (UTC) — fixes the midnight-to-02:00 off-by-one
  that truncated a day from YTD / this-month / this-quarter for Swedish
  users.
- NoDocRequiredToggle restores the previous reason on failed POST/DELETE so
  the rolled-back toggle state stays consistent with the rendered reason.
- Document the company-scoped (not user-scoped) DELETE authorization policy
  on the no-document-required route.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-28 01:56:09 +02:00
Mattsson 627109b5bd Mcp/entry bug (#580) 
* feat: restore document version supersession on posted entries

* refactor: update VAT handling logic to remove dependency on seller registration status

* feat: update VAT handling to respect seller registration status across invoice processing

* feat: add no-op migration placeholder for document supersession
 2026-05-27 17:07:27 +02:00
Mattsson 8979f6eda3 Bug/year end failure (#575) 
* feat: implement findNextPeriod function and integrate into year-end closing logic

* feat: add integrity check for PDF documents and enhance user feedback for corrupt files

* Refactor year-end service and period creation logic for improved UTC handling and error messaging

- Update `validateYearEndReadiness` to assert on stable warning messages without interpolating period names.
- Modify `createNextPeriod` to ensure date calculations are performed in UTC, preventing DST-related issues.
- Enhance error handling in `validateYearEndReadiness` and `executeYearEndClosing` to avoid exposing database details.
- Introduce structured error messages for year-end processes in `structured-errors.ts`.
- Add tests for document integrity checks, ensuring proper authentication and error handling.
- Implement GUC checks in document versioning to prevent unauthorized modifications and ensure company membership.
- Update migration scripts to reflect changes in document immutability enforcement.

* fix: add comment to clarify GUC behavior in document supersession logic
 2026-05-27 14:19:30 +02:00
Mattsson a2a556d837 Bug/UI wrong display (#573) 
* fix(dashboard): exclude credit notes from unpaid invoices widget

Credit notes (status='sent', negative total) were summed into the
"Att få betalt" widget, producing confusing negative totals like
"2 st, -38 625 kr". Filter them out via credited_invoice_id IS NULL,
matching the existing pattern in reminder-processor and the AR ledger.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(documents): harden PDF preview and upload validation

- JournalEntryAttachments: switch inline PDF preview from <iframe> to
  <object type="application/pdf">. Mirrors the AttachmentPreviewSheet
  fix from #572 — Chrome's frame pipeline intermittently surfaced
  "Det här innehållet har blockerats" on iframes even with permissive
  CSP. <object> invokes the PDF plugin directly. crbug.com/271452.

- /api/documents/:id/inline: resolve Content-Type via file extension
  when mime_type is null or application/octet-stream. Legacy uploads
  landed with empty File.type from some drag sources; combined with
  the new X-Content-Type-Options: nosniff header on this route,
  Chrome refused to render valid PDFs. Extension fallback covers
  every legacy row without a DB backfill.

- /api/documents POST: surface DB-trigger period-lock errors as a
  400 DOC_UPLOAD_PERIOD_LOCKED with a Swedish reason. Previously
  every catch was bucketed into DOC_UPLOAD_STORAGE_FAILED (500 /
  "Filen kunde inte sparas") which hid the real cause from users
  attaching to verifikationer in closed/locked fiscal periods.

- document-service: add validateDocumentMagicBytes() that inspects
  the first bytes for valid PDF/PNG/JPEG/WebP headers (PDF tolerates
  a leading UTF-8 BOM). Wired into uploadDocument() and
  createNewVersion() so every upload path is protected — UI, MCP,
  and future email/webhook ingestion. Defends against agents that
  send a base64-encoded text placeholder instead of real binary
  bytes via the gnubok_upload_document MCP tool, which produced
  tiny (15-561 byte) "PDFs" that failed to render in Chrome and
  in external viewers.

Tests use a minimal valid PDF buffer (%PDF-1.4 … %%EOF).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(arsredovisning): emit ÅRL-required notes and FTE-weighted medelantal

Five compliance gaps fixed in the K2 and K3 noter builders:

- Anläggningstillgångar roll-forward per ÅRL 5:8 § — per-category IB
  anskaffningsvärde, tillkommande, avgående, UB and accumulated
  avskrivningar movement (was only emitting avskrivningstider).
- Långfristiga skulder förfallande efter mer än fem år per ÅRL 5:13 §.
- Ställda säkerheter and Eventualförpliktelser as separate notes per
  ÅRL 5:14-15 § (K2 previously combined them).
- Koncernförhållanden per BFNAR 2016:10 kap. 19 / BFNAR 2012:1 kap. 8.

Replaces medelantal anställda — the old query filtered employees by an
is_active column that doesn't exist, so the note never emitted. Now
uses an FTE-weighted day-based average per ÅRL 5:20 §.

Six disclosure fields persist on arsredovisning_narratives as per-period
overrides; the UI extends the existing förvaltningsberättelse editor
with a "Lagstadgade upplysningar" subsection sharing the same Spara
button — no new pages, no settings changes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(invoices): respect vat_registered=false and hide personnummer for B2C

- PDF address block no longer prints org_number for individual customers
  (GDPR data minimization; ML 17 kap 24§ requires name + address only).
- Wire company_settings.vat_registered through the rule helpers, invoice
  creation API, preview-pdf API, and the new-invoice form so a non-VAT-
  registered seller cannot charge VAT (ML 1 kap. 1§). The PDF suppresses
  the empty "Moms 0%" row and shows a dedicated "Företaget är inte
  momsregistrerat" notice instead of the ML 3 kap. exempt notice.
- Engine unchanged: 'exempt' treatment already routes to 3004/3100 and
  skips VAT lines.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(settings): remove approval rules from sidebar and routes

* fix(invoices): ensure vat_registered defaults to true for invoice previews and API

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-27 11:01:53 +02:00
Mattsson 32d9978f1b Fix/chrome pdf preview csp (#572) 
* feat: add option to exclude year-end closing entries in SIE export and related reports

* delete docs

* fix: allow Chrome's PDF viewer in verifikat document preview

The /api/documents/:id/inline route shipped with
`object-src 'none'` in its CSP, which blocked Chrome's built-in PDF
viewer (it renders inline PDFs via an internal <embed>). Users on
Chrome saw "Det här innehållet har blockerats" when expanding a PDF
attachment in the bookkeeping view; Firefox (PDF.js) and Edge (own
viewer) were unaffected, and JPGs worked because <img> isn't subject
to object-src.

Drops the CSP for this route to the minimum needed for embeddability:
`frame-ancestors 'self'`. X-Content-Type-Options: nosniff plus the
fixed Content-Type from the handler already block MIME confusion;
X-Frame-Options: SAMEORIGIN + frame-ancestors still block clickjacking.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(auth): add webmail deep link to email confirmation screens

Mirrors Stripe's signup UX: after asking the user to verify their email,
detect their webmail provider from the domain and show a button that
opens the inbox in a new tab. Gmail gets a from:<sender> search
pre-populated; Outlook/Yahoo/iCloud/Proton open the inbox directly.
Unknown / custom domains fall back to the existing copy.

Sender address is configurable via NEXT_PUBLIC_BRANDING_AUTH_EMAIL_FROM
(default noreply@gnubok.se) so white-label installs can match their
Supabase Auth SMTP config.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(auth): unblock first-time password set for BankID users with MFA

Supabase rejects updateUser({password}) and mfa.unenroll with "AAL2 session
is required" whenever a TOTP factor is enrolled. BankID magic-link logins
produce AAL1, and middleware skips MFA enforcement for bankid_linked users,
so they had no path to AAL2 — leaving them unable to set a backup password
or disable MFA without going through the email-recovery escape hatch.

- /api/account/password: branch on app_metadata.has_password. First-time set
  writes via service.auth.admin.updateUserById (no existing credential to
  protect, AAL2 guard does not apply). Change-password keeps the user-session
  updateUser so AAL2 still fires for credential rotation.
- /mfa/verify: accept a safeReturnTo query param and route there after
  successful verify, so step-up flows can land back where they came from.
- SecuritySettings: detect the AAL2 error from both change-password and
  mfa.unenroll and redirect through /mfa/verify?returnTo=/settings/account
  instead of toasting a dead-end error.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* Add tests and rounding utility for öre precision in bokslut calculations

- Implemented `roundOre` function for rounding SEK amounts to two decimal places, ensuring consistent monetary calculations.
- Introduced `ORE_TOLERANCE` constant for comparing rounded amounts, facilitating invariant checks in financial entries.
- Created comprehensive tests for `roundOre`, covering typical cases, edge cases, and idempotency.
- Added year-end invariants tests to verify database-level guarantees for closing entries, ensuring they balance to the öre and reject discrepancies.
- Developed end-to-end tests for the dispositions chain, validating the correctness of calculations across various scenarios.

* fix: update PDF rendering to remove Swish QR code generation and set default to disable Swish visibility

* fix: enhance security by rejecting data URIs in safeReturnTo function tests

* fix: improve rounding logic in roundOre function and add customer_type migration

* fix: add customer_type column to customers and enforce CHECK constraint

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-26 22:29:41 +02:00
Mattsson 78c91e00e4 feat: add language preference for customers to support invoice locali… (#561) 
* feat: add language preference for customers to support invoice localization

- Introduced language support for invoices, allowing customers to choose between Swedish and English.
- Updated invoice PDF generation to reflect the selected language for titles, labels, and messages.
- Enhanced email templates to generate content in the customer's preferred language.
- Added migration to include a language column in the customers table with a default value of Swedish.
- Updated tests to verify correct language usage in invoice emails and PDFs.

* fix: debounce API requests in InvoicePreviewCard and update F-skatt terminology in email templates
 2026-05-22 15:21:05 +02:00
Mattsson 64bbeb4021 Fixed user issues (#559) 
* Fixed user issues

* feat: add personal_number column to customers for individual identification

* feat: add personal_number field to makeCustomer function for enhanced customer identification

* feat: add personal_number column with constraint check for customer identification
 2026-05-22 12:22:15 +02:00
Mattsson b34de598e3 feat: add inbox-direct supplier invoice creation from inbox items (#558) 
* feat: add inbox-direct supplier invoice creation from inbox items

- Implemented `gnubok_create_supplier_invoice_from_inbox` tool in the MCP server for creating supplier invoices directly from inbox items.
- Enhanced the input schema to include `inbox_item_id` and `document_id` for direct booking.
- Added logic to validate inbox items and link documents to journal entries during the commit process.
- Introduced `commitCreateSupplierInvoiceFromInbox` function to handle the creation and linking of supplier invoices.
- Added unit tests to cover various scenarios including happy path, idempotency, error handling, and rollbacks.
- Updated database migration to extend the `pending_operations` table to include the new operation type.

* fix: extend CHECK constraint to include create_supplier_invoice_from_inbox operation

* feat: add validation for financial fields in supplier invoice creation from inbox
 2026-05-22 11:09:46 +02:00
Jakob Wennberg f8f49f8426 Inbox page-count gate + DataList/DropdownMenu primitives (#554) 
* feat(inbox): skip AI extraction for multi-page PDFs (#553)

Bedrock churns for minutes on multi-page PDFs (sales reports, bank
statements, contracts) and returns nothing useful. Above 3 pages we now
skip extraction entirely and mark the row with extraction_skipped=true;
the document still lands in the inbox and can be attached or converted
manually. Same gate applies to the /items/:id/attach path. Client can
also opt out via skip_extraction=true (skip_reason=client_opt_out).

The InvoiceInboxWorkspace renders an "Inte AI-tolkad" badge for skipped
rows, distinct from the "Felaktig" failure state (status='error').

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* refactor(ui): introduce DataList + DropdownMenu primitives, roll out across list pages

DataList replaces the per-row Card pattern across Granskning, Transactions,
Invoices, Supplier invoices, and Pending. One bordered container with
hairline rows matches the flat-with-hairlines aesthetic in CLAUDE.md —
no shadows, no state-tinted borders, secondary token for selected/hover.

DropdownMenu fills the gap for row-level action menus on TransactionInboxCard,
TransactionHistoryList, and the page-level action menus on /transactions and
/pending. Replaces ad-hoc Popover + buttons constructions.

Migrates list pages and the transaction inbox/history components onto the
new primitives. No behavior change beyond the visual unification.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* chore: add agent skills + gnubok domain skills, gitignore compliance reports

.agents/skills/ + skills-lock.json + symlinks under .claude/skills/ check in
the vercel-labs/agent-skills set pinned by the local skill manager
(deploy-to-vercel, vercel-cli-with-tokens, react-best-practices,
composition-patterns, react-native-skills, react-view-transitions,
web-design-guidelines). Keeps the team on the same versions.

.claude/skills/industry/ + .claude/skills/modifier/ are hand-authored
vertical and entity-modifier skills for the specialized accountant agent
— industries (konsult-it, e-handel, bygg-hantverk, reklambyra, saas-ai)
and entity overlays (holding-ab, single-shareholder-ab-fmb, mixed-
verksamhet). Project-owned content; lives in the repo by design.

Also gitignores .compliance-reports/ — those are large generated
SARIF/dossier artifacts from the compliance scanner.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* refactor(transactions): unify inbox/history chrome, drop swipe flow

The transactions page mixed two-tier filtering, a swipe-view detour, and
per-row Card chrome that didn't carry its weight. This pass collapses
those into a single editorial-list surface and removes the unused swipe
path entirely.

User-visible changes:
- Mode toggle (Att bokföra / Alla transaktioner) moved from a Tabs row
  under the header into a dropdown to the right of a unified search bar.
  Search now persists when switching modes.
- Removed the swipe categorization view ("Gå igenom alla") and its
  trigger button. The 800-line SwipeCategorizationView component is
  deleted; suggestion-fetching shrinks to what the template picker
  still consumes.
- Inbox rows now show one primary action: invoice/supplier-invoice
  match shortcut when auto-detected, else "Bokför". A new visible Link2
  icon button opens the customer or supplier invoice picker manually
  (chosen by amount sign). Delete becomes a plain trash button — no
  overflow menu since it only ever held one item.
- Bulk action bar swaps "Markera som privat" for "Ta bort" with a
  single combined confirmation.
- Built SupplierInvoicePicker mirroring InvoicePicker so expense
  transactions can be matched to supplier invoices from the inbox.
  Wired through /api/transactions/{id}/match-supplier-invoice.
- Template picker dialog renamed to "Bokför transaktion"; "Bokför
  manuellt…" and "Matcha med faktura…" promoted from muted ghost
  buttons at the bottom to outline buttons at the top, above the
  template list.
- Breathing room: row padding py-3 → py-4, primary text text-sm →
  text-base, amount text-base, button heights h-8 → h-9, trailing
  gap-2 → gap-3 (in the DataList primitive itself, so every list
  benefits slightly).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix: align package-lock.json with merged package.json

The merge resolution took origin/main's package-lock.json (which dropped
pdf-lib) but kept our package.json (which still requires pdf-lib for the
invoice-inbox extension's PDFDocument import). `npm ci` rejected the
mismatch.

Regenerate the lock from the merged package.json so both files agree.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix: regenerate package-lock.json with npm@10 for CI compat

Local npm@11 produced a lock that npm@10 (CI) rejected with
"Missing: @swc/helpers@0.5.21". Regenerated with npm@10
--package-lock-only so CI can install.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(transactions): supplier-invoice status-leak guard + drop dead prop

Two follow-ups from the merge-risk audit:

- SupplierInvoicePicker now mirrors InvoicePicker's status-leak guard:
  if a supplier invoice is still 'approved'/'overdue' but already has a
  payment voucher attached (journal_entry_id on supplier_invoice_payments),
  hide it. Closes a UX race window between payment and status flip.
  Partially-paid invoices still pass through.
- Drop the unused onMarkPrivate prop on TransactionInboxCard and the
  matching handleMarkPrivate wrapper in the parent. Both became dead
  when the swipe-categorisation flow was removed.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-22 10:12:21 +02:00
Mattsson e4488a900b feat: add user locale preference to user_preferences table (#555) 
* feat: add user locale preference to user_preferences table

- Introduced a new column 'locale' in the user_preferences table to store per-user UI language preferences.
- Added a CHECK constraint to ensure only supported locales ('sv', 'en') are allowed.
- Triggered a schema reload notification for the changes.

chore: declare CSS module support in TypeScript

- Added a declaration for CSS modules in globals.d.ts to enable TypeScript support for importing CSS files.

* feat: add Swish as an invoice payment method in company settings
 2026-05-21 21:33:22 +02:00
Jakob Wennberg cc351158f8 Invoicing & account-security polish bundle (#550) 
* feat: invoicing & account-security polish bundle

Five independent improvements bundled to ship together:

- BankID/password lockout fix: BankID-only users could enroll MFA and
  brick themselves (Supabase requires AAL2 to change password or unenroll
  MFA, and AAL2 needs a password sign-in). New app_metadata.has_password
  flag tracks this; middleware gates /mfa/enroll behind it, /account/set-
  password is the unlock path, SecuritySettings shows a banner, and
  /api/account/password is the single write path that flips the flag.
  Backfill script for existing users.

- Swish invoice payment method: company_settings.swish + invoice_show_swish
  columns, validation in lib/api/schemas.ts (accepts 123XXXXXXX företag or
  07XXXXXXXX mobile, strips whitespace/hyphens), rendered on invoice PDFs.

- Send-reminders kill switch: per-company company_settings.send_invoice_
  reminders toggle in PdfPrintSettings/Automatisering. Reminder processor
  also tightened: positive status allowlist (sent + overdue) so terminal
  statuses can never match; skip when customer already responded via
  reminder link; race-window re-check before send.

- First-invoice logo prompt: one-shot dialog when creating the first
  invoice without a logo (issue #520). Self-limits via head-only count.

- SIE export opening-balance fallback: route IB through getOpeningBalances
  so the compute_prior_opening_balances RPC supplies #IB after multi-year
  imports where opening_balance_entry_id is intentionally NULL. Previously
  #IB silently went to zero and #UB collapsed to current-period movements.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(account-polish): address PR review feedback

- BankID-link path (extensions/general/tic/index.ts): read-merge-write
  app_metadata instead of passing { bankid_linked: true } alone.
  updateUserById REPLACES app_metadata wholesale, so the previous code
  would have wiped has_password for any user who later linked BankID,
  causing the set-password banner to (incorrectly) reappear and blocking
  the standard MFA enrollment button. The comment is now corrected.

- Middleware (lib/supabase/middleware.ts): thread inner returnTo through
  the /mfa/enroll → /account/set-password redirect so the user lands on
  their original destination after the full chain completes, not on /.

- safeReturnTo helper (lib/auth/safe-return-to.ts): replace the
  starts-with-/-but-not-// guard on mfa/enroll and set-password pages.
  The previous guard let /\evil.com and /@evil.com through. The new
  helper parses against a synthetic base origin and verifies it matches.

- set-password page (app/(auth)/account/set-password/page.tsx): remove
  CLAUDE.md design system violations — bg-gradient-to-b on page bg,
  inline shadow-md style on the card, space-y-5, font-medium on the h1,
  rounded-xl on the card. Flat surface, hairline border, font-display
  h1 per the design tokens.

- Swish dedup (lib/payments/swish.ts): extract normaliseSwish() and
  isValidSwish() helpers and use them in lib/api/schemas.ts,
  components/settings/BankDetailsForm.tsx, and the invoicing settings
  page. Single source of truth for the regex.

- Password route (app/api/account/password/route.ts): emit a structured
  success log so the audit pipeline can detect password-set events, not
  just failures.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-21 16:44:09 +02:00
Mattsson e71b4a9138 Add/mcp and visma (#547) 
* fix: simplify COMING_SOON_PROVIDERS to include only bjornlunden and briox

* feat: add supplier creation functionality and related operations

* feat: reorder and enhance OAuth scopes in Visma integration

* feat: implement create supplier functionality with validation and risk tier management
 2026-05-21 01:25:18 +02:00
Mattsson 9aced4790c feat(api): implement caching and logging in health check endpoint (#526) 
* feat(api): implement caching and logging in health check endpoint

- Added in-memory caching for health check responses to reduce load on Postgres.
- Introduced logging for error handling in health check.
- Updated response structure to exclude error details from public responses.

feat(api): enhance OAuth consent UI and scope handling

- Improved consent UI to reflect exact requested scopes and added better user guidance.
- Updated scope handling logic to ensure least-privilege access.
- Enhanced styling for better user experience and accessibility.

chore(docker): improve security and resource management in Docker setup

- Updated Docker Compose configuration to enforce read-only file systems and resource limits.
- Added health checks and logging options for better observability.
- Introduced optional Caddy reverse proxy for TLS termination.

fix(migrations): resolve ambiguity in create_company_with_owner function

- Dropped orphaned 3-arg overload of create_company_with_owner function.
- Recreated canonical 4-arg version with cash account seeding logic.
- Ensured proper permissions for function execution in Postgres.

* feat: enhance security checks for team membership in company creation
 2026-05-19 17:52:11 +02:00
Mattsson 8a6ce7093e feat: implement skattekonto drift detection and alerting (#525) 
* feat: implement skattekonto drift detection and alerting

- Add skattekonto drift computation logic to compare Skatteverket's saldo with GL 1630 sum.
- Implement alerting mechanism for significant drift changes, with throttling to prevent alert spamming.
- Introduce database functions to sum GL 1630 entries and list unbooked skattekonto rows.

feat: create own account transfer detection

- Develop logic to detect transfers between a company's own cash accounts based on counterparty IBAN.
- Implement tests to validate detection logic under various scenarios, including matching and non-matching IBANs.

feat: establish cash accounts as a first-class entity

- Create cash_accounts table to manage routable cash accounts, replacing ad-hoc JSONB structures.
- Implement functions for listing, upserting, and managing cash accounts, including primary account designation.

feat: enhance GL line reconciliation functionality

- Modify get_unlinked_1930_lines RPC to accept any account number for reconciliation, improving flexibility for different currencies.
- Update related functions to ensure compatibility with the new cash_accounts structure.

feat: capture counterparty IBAN in transactions

- Add counterparty_iban column to transactions table to facilitate intra-account transfer detection.
- Create index for efficient lookups based on counterparty IBAN.

* feat: Enhance cash account handling and reconciliation processes

- Updated reconciliation routes to enforce cash account validation for all account numbers, including '1930'.
- Improved error handling for unknown cash accounts in reconciliation status and unmatched entries routes.
- Changed CashAccountSelector to use sessionStorage instead of localStorage for better data privacy.
- Fixed mapping for employer payroll taxes to route to the correct account (2730 instead of 2731).
- Added safety checks for company IDs in the guessCounterAccount function to prevent injection vulnerabilities.
- Introduced atomic RPC for setting primary cash accounts to avoid intermediate states during updates.
- Seeded default cash accounts for new companies to ensure reconciliation routes are accessible from day one.
- Updated email notifications for drift detection to avoid exposing sensitive financial data.
- Enhanced bank reconciliation logic to handle multi-currency transactions correctly.
- Renamed and updated tests to reflect changes in the underlying RPCs and ensure accurate coverage.
- Migrated existing cash account rules to correct mappings in compliance with Swedish accounting standards.
 2026-05-19 16:10:18 +02:00
Mattsson e211ab31be UI/settings api mcp (#524) 
* feat(voucher): add create voucher and correct entry previews; update commit methods

* feat: add support for pending operations in API key scopes and OAuth client management

- Introduced new API key scopes for reading and approving pending operations.
- Updated the scope groups to include pending operations.
- Added new tools for listing and managing pending operations.
- Implemented OAuth client registration and revocation endpoints.
- Created a UI panel for managing OAuth clients, including registration and revocation.
- Added tests for pending operations tools and OAuth allowlist functionality.
- Implemented a database migration for OAuth client registrations with appropriate policies and constraints.

* feat: Implement OAuth client registration rate limiting and enhance security measures

- Added IP-based rate limiting to the OAuth client registration endpoint to prevent enumeration attacks.
- Introduced a service-role client for allowlist lookups, ensuring trust boundaries are maintained.
- Updated error responses to be uniform across different types of redirect URI validation failures.
- Enhanced tests to reflect changes in OAuth scope handling, ensuring fallback to read-only scopes when no scopes are provided.
- Improved handling of high-risk pending operations, requiring explicit confirmation for approvals.
- Added audit logging for OAuth client revocations and pending operation approvals/rejections to maintain a security audit trail.
- Refactored API key scope management to include default read-only scopes for OAuth-issued keys and added segregation-of-duties checks.

* feat: add recurring invoice scheduling functionality

- Implemented recurring invoice schedules with a new database schema.
- Created API routes for managing recurring invoices (GET and POST).
- Added cron job to automatically generate invoices based on schedules.
- Developed service functions for computing next run dates and executing schedules.
- Added tests for the new functionality, including validation and success cases.
- Introduced error handling for various scenarios in the invoice creation process.

* feat: refine VAT rate validation and enhance recurring invoice handling
 2026-05-19 13:48:32 +02:00
Mattsson 16164ea14c Fix/mcp fixes and bugs (#518) 
* feat(voucher): add create voucher and correct entry previews; update commit methods

* feat: add support for pending operations in API key scopes and OAuth client management

- Introduced new API key scopes for reading and approving pending operations.
- Updated the scope groups to include pending operations.
- Added new tools for listing and managing pending operations.
- Implemented OAuth client registration and revocation endpoints.
- Created a UI panel for managing OAuth clients, including registration and revocation.
- Added tests for pending operations tools and OAuth allowlist functionality.
- Implemented a database migration for OAuth client registrations with appropriate policies and constraints.

* feat: Implement OAuth client registration rate limiting and enhance security measures

- Added IP-based rate limiting to the OAuth client registration endpoint to prevent enumeration attacks.
- Introduced a service-role client for allowlist lookups, ensuring trust boundaries are maintained.
- Updated error responses to be uniform across different types of redirect URI validation failures.
- Enhanced tests to reflect changes in OAuth scope handling, ensuring fallback to read-only scopes when no scopes are provided.
- Improved handling of high-risk pending operations, requiring explicit confirmation for approvals.
- Added audit logging for OAuth client revocations and pending operation approvals/rejections to maintain a security audit trail.
- Refactored API key scope management to include default read-only scopes for OAuth-issued keys and added segregation-of-duties checks.
 2026-05-18 19:02:42 +02:00
Mattsson a652dcae1a Skv/e2e overview (#515) 
* feat(agi): refactor AGI XML generation and data handling

- Remove deprecated AGI field codes from field-codes.ts.
- Update generate-declaration.ts to include new employee fields and handle absence data with stable specification numbers.
- Enhance XML generation in xml-generator.ts to support new flags for housing benefits and adjusted benefits.
- Introduce new database migrations to support:
  - `removed_from_agi` flag for tombstoning individuppgifter.
  - `benefits_adjusted` flag for tracking adjustments to benefits.
  - `franvaro_specifikationsnummer` for stable absence event identification.
  - `housing_benefit_type` to differentiate between housing benefit types.

* feat: Implement strict validation for AGI employee data and introduce pre-flight validation schemas

- Added Zod schemas for validating employee data in AGI declarations to ensure all required fields are present and correctly typed, preventing silent errors during processing.
- Introduced AGI pre-flight validation schemas for Skatteverket endpoints to validate individual and head unit submissions before sending to the API.
- Created a new audit log table for tracking all outbound calls to Skatteverket, ensuring compliance and traceability for AGI and moms submissions.
- Implemented advisory locks in the database to manage concurrent updates to absence specification numbers, enhancing data integrity.
- Added compliance documentation for GDPR processing activities related to AGI and moms submissions, detailing data handling and retention policies.

* feat: Extend DELETE RLS policy to protect 'declined' signatures in årsredovisning

* fix: Update date handling in salary absence migrations to use immutable year-month key

* fix: Update SELECT policy in skatteverket_api_audit_log to use IN clause for company_id

* feat: Add skatteverket_api_audit_log and salary_absence_franvaro_audit tables with RLS policies and immutable triggers
 2026-05-17 23:52:49 +02:00
Jakob Wennberg 831920fede fix(arcim): allow Fortnox re-sync to replace prior SIE import per fiscal year (#512) 
* fix(arcim): allow Fortnox re-sync to replace prior SIE import per fiscal year

A user reported a sync failure when retrying Fortnox after adding more
verifications:

  Import failed: Failed to create pending import record:
  duplicate key value violates unique constraint
  "sie_imports_company_id_file_hash_active_idx"

Root cause: Fortnox embeds the export-time #GEN date in every SIE export,
so the file hash always differs between syncs. The wizard's hash-based
duplicate detection treated each sync as a brand new file, but the engine
still rejected the insert because the per-period import slot was held by
the prior 'completed' row.

This change reframes Fortnox re-sync as a replace operation rather than a
fresh import:

- New executeSIEImport option `onExistingPeriod: 'block' | 'replace'`.
  Manual SIE upload at /api/import/sie keeps default 'block' (current
  behavior, no regression). The Fortnox /import-sie endpoint passes
  'replace', which runs replaceSIEImport on any overlapping completed
  import before insert. Imported journal entries from the prior import are
  cancelled per BFL 5 kap 5§; user-created entries (manual, transaction,
  invoice) are untouched.

- /sie-data switches from hash-based to period-based duplicate detection
  and returns previousImport metadata per fiscal year.

- Wizard drops the alreadyImported skip filter, surfaces an amber callout
  in the confirm dialog listing fiscal years that will be replaced, and
  shows "ersatte N tidigare importerade verifikationer" per year.

- createPendingImportRecord translates 23505 partial-index violations to
  a clear Swedish recovery message instead of leaking the raw constraint
  name.

- cleanupStaleImportRecords drops the 1-hour age gate and also cleans
  status='mapped' orphans. SIE imports are single-flight per company so
  the gate just made legitimate retries fail.

- After replace, the fiscal_periods row's opening_balances_set and
  opening_balance_entry_id are cleared (only when they pointed at the
  cancelled prior IB entry), so the new IB import isn't skipped.

Schema-drift migration captures the partial unique index
sie_imports_company_id_file_hash_active_idx that already exists in
production (added out-of-band) and drops the now-superseded plain
sie_imports_company_id_file_hash_key constraint. Both statements are
idempotent — verified no-op against production.

Tests: new pg-real test covers the partial index admit-replaced
semantics, source_type='import'-only cancellation in replace_sie_import,
and the post-replace insert path.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(arcim): address PR review — restore 5-min cleanup gate, fix test source_type

Greptile P2: `cleanupStaleImportRecords` was deleting `pending` rows
unconditionally, which could wipe a concurrent in-flight import in
another tab/session. Restored a 5-minute age gate (long enough for any
normal interactive import, short enough that legitimate retries after
a crash still succeed). Also dropped `mapped` from the cleanup — it is
defined in SIEImportStatus but no code path writes it, so including it
was both unnecessary and added the concurrent-session risk Greptile
flagged.

pg-real test: insertPostedEntry used `source_type='transaction'` which
is not a valid value per the journal_entries_source_type_check
constraint (migration 20260516060000). Switched to `'bank_transaction'`
— the actual source_type emitted when a user categorizes a bank
transaction in gnubok.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-17 16:25:43 +02:00
Jakob Wennberg f98ffee145 feat(bokslut): Phase 8 — make årsredovisning Bolagsverket-fileable (#511) 
* feat(bokslut): Phase 8 — make årsredovisning Bolagsverket-fileable

Closes most of the deferred items from PR #509 review. The ÅR PDF is now
an honest draft a user can take to Bolagsverket: it includes the fastställe-
intyg page ÅRL 8 kap 3 § requires, the K2-mandatory aktiekapital note, and
the narrative edits actually survive a refresh.

Narrative persistence (replaces the round-1 URL-query-param carry)
- New table arsredovisning_narratives (UNIQUE per fiscal_period_id, length
  caps matching the API schema, RLS + updated_at trigger).
- narrative-service.ts: getNarrative / upsertNarrative.
- /api/.../arsredovisning/narrative GET + POST. POST does an explicit
  period-ownership pre-check before the upsert.
- buildArsredovisningData loads persisted narrative as override layer
  (caller-supplied overrides → persisted → boilerplate).
- ÅR page replaces the URL-query-param hack with a Spara button + saved
  indicator. The PDF download URL is plain again — no narrative content
  in access logs, browser history, or CDN logs.
- PDF route stops parsing description/events/disposition query params.
  Also closes the GDPR Art.25(1) finding the bot flagged in PR #509.

Fastställelseintyg PDF page
- New 7th page in ArsredovisningPDF after Underskrifter. Carries the ÅRL
  8 kap 3 § attestation text + the resultatdisposition + a signature slot.
- Without this page Bolagsverket rejects the filing — flagged in the round-2
  Swedish review on PR #509.

K2 aktiekapital note + framework guard
- buildK2Noter now takes entityType. Note 1 only claims K2 when the
  company is an AB; non-AB gets a generic principles statement so we
  don't falsely assert a framework. Future K3 election will flip this
  branch when it lands.
- New aktiekapital note (required K2 note for AB per BFNAR 2016:10 ch.18).
  Reads aktiekapital / antal_aktier / kvotvärde from company_settings;
  emits a "saknas — komplettera under Inställningar" placeholder when
  missing.

Manual "Mark as signed" PATCH + UI button
- New PATCH /signatures/[signatureId] — flips pending → signed (manual
  / paper flow) or pending → declined. Real BankID wiring is Phase 9 and
  will use the same markSignatureSigned helper with the BankID callback
  as the trigger.
- ÅR page renders a "Markera som signerad" button on every pending row.

Small cleanups all flagged in PR #509 reviews
- AccrualProposal.reverses_on type: '' → null. The future accrual-reversal
  cron will filter `reverses_on IS NOT NULL`; an empty string would
  silently match.
- ArsredovisningData.company.sate → city. The typo carried into the type
  in earlier phases; renaming now before any external consumer takes a
  dependency.
- signer_name CHECK length 200 at storage layer (matches the API .max(200)
  added in PR #509 round-2 — GDPR Art.25.2 belt-and-braces).
- Soliditet equity filter now has a code comment explaining the K2 vs K3
  branch the bot wanted documented for the future K3 migration.

Explicit follow-ups (each merits its own focused PR):
- Real BankID signing — needs provider choice + polling + QR. Phase 9.
- Accrual reversal cron — auto-flip 17xx/29xx accruals on Jan 1 of next FY.
- Medelantal anställda proper annual average — needs salary-run aggregation.
- Vacation avgifter age-tier split (10.21 % for 67+) — needs upstream
  vacation-liability report to expose age.
- K2 noter expansion (lån till närstående, eventualförpliktelser detail).

Verification
- 94 unit tests pass (bokslut + MCP subsets)
- Zero typecheck errors on any touched file
- Zero lint errors on any touched file
- Migration 20260517140000 applied to remote Supabase via MCP

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bokslut): address PR #511 round-1 — 3 P1s + 3 real concerns

3 P1s from Greptile (all real bugs):

- entityType default reintroduced the K2 false-assertion. build-data.ts
  defaulted `entity_type ?? 'aktiebolag'`, which means every unconfigured
  company would still claim K2 in Note 1 — exactly the false-assertion the
  framework guard was added to prevent. Now defaults to 'unknown' and the
  guard treats that as not-K2. New warning surfaces in the data so the UI
  can prompt the user to fill in företagsform.

- Signatures PATCH ignored the URL fiscal-period id. The route destructured
  `id` from params but never used it as a filter, so PATCH /periods/A/
  signatures/SIG_FROM_B succeeded silently — broken REST contract + IDOR
  across periods. Rewrote the handler to do a single UPDATE with all four
  filters: id, company_id, fiscal_period_id, status='pending'. Missing row
  returns 409 SIGNATURE_INVALID_TRANSITION instead of silent 200.

- Signatures state-machine guard was missing. Without status='pending' in
  the WHERE clause, an already-signed signature could be flipped back to
  declined (or vice-versa). Now part of the consolidated UPDATE above.

3 real concerns:

- Narrative GET lacked ownership pre-check. POST already had it; mirroring
  on GET so a valid JWT for company A can't probe / enumerate company B's
  period IDs through the narrative endpoint.

- Narrative POST lacked period-lock check. BFL 5 kap 5 § makes
  räkenskapsinformation immutable after filing — editing the
  förvaltningsberättelse on a closed/locked period now returns
  PERIOD_LOCKED.

- Aktiekapital placeholder text would land in Bolagsverket-filed PDF body.
  When aktiekapital fields are missing, the note now omits entirely and a
  warning surfaces in the ArsredovisningData.warnings array — the UI flags
  it pre-download with a "Innan inlämning till Bolagsverket" list. Same
  surface picks up the entityType=unknown and entityType=non-AB warnings.

Plus 2 schema improvements from Swedish review:

- AGM date persistence. Fastställelseintyg date was a literal "____" blank,
  defeating the point of a generated PDF. New agm_date column on
  arsredovisning_narratives + UI date input + PDF now renders the saved
  date. When missing, the warning surface flags it.

- Composite UNIQUE constraint on (company_id, fiscal_period_id) instead of
  just fiscal_period_id. UUIDs don't collide across tenants in practice
  but the constraint should match the tenant boundary so a logic error in
  onConflict resolution can't write to another company's row. Migration
  20260517160000 drops the old constraint and adds the composite.

Verification
- 94 unit tests pass
- Zero typecheck errors on any touched file
- Zero lint errors on any touched file
- Migration 20260517160000 applied to remote Supabase via MCP

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bokslut): address PR #511 round-2 — 5 real concerns + BFL/GDPR conflict

5 real concerns from the round-1 bot re-eval:

- Narrative SELECT * leaked user_id to the frontend. getNarrative and
  upsertNarrative now project an explicit NARRATIVE_API_COLUMNS list
  (id, company_id, fiscal_period_id, narrative fields, agm_date, updated_at).
  user_id and created_at stay server-side. NarrativeRow type updated to
  match. Closes Art.25.2 + 2× A.8.3.

- agm_date validated only as YYYY-MM-DD regex. '2024-13-99' passed Zod
  and surfaced as a Postgres 500 instead of a 400. Added a refine() that
  parses with new Date() and confirms ISO round-trip equality, so invalid
  calendar dates return a clean structured-error.

- agm_date had no range check. ÅRL 8:3 → 7:10 §§ requires the AGM to be
  held after period end and within 6 months for privat AB; build-data
  warnings now flag agm_date <= period_end (impossible) and agm_date >
  period_end + 6 months (deadline). Warning surface in the UI already
  picks these up from the existing list.

- Fastställelseintyg signer label "Styrelseledamot / VD" conflated
  legally distinct roles per ÅRL 8:3 → 6:6-7 §§ — a VD without board
  membership cannot sign. Label is now "Styrelseledamot (närvarande vid
  stämman)" and the body text references the AGM's resolution
  ("stämmobeslutet") rather than the board's proposal — the AGM votes,
  and it is the vote that must be certified.

- Aktiekapital warning suppressed for entityType='unknown'. The maybeAb
  branch in buildK2Noter now fires for both 'aktiebolag' and 'unknown'
  so an unconfigured company that's actually an AB still gets prompted
  to fill in aktiekapital before filing. Note body stays omitted when
  fields are missing; only the warning surfaces.

BFL × GDPR conflict (new migration 20260517180000):

- Both arsredovisning_narratives and arsredovisning_signature_requests had
  user_id with ON DELETE CASCADE → auth.users. BFL 7 kap 1 § requires
  räkenskapsinformation to be retained for 7 years; GDPR Art.17 erasure
  or membership revocation would silently delete filed årsredovisning
  narrative + BankID signature evidence. BFL wins for filed financial
  records — user_id is now nullable with ON DELETE SET NULL on both
  tables. The company FK keeps its CASCADE (company deletion takes its
  räkenskapsinformation with it; that's a separate workflow).

Deliberately not chasing on this round:
- ISO A.8.12 historical PDF query-param logs — process item for the risk
  register, not code (the leak path is closed in this PR's first commit).
- "Collapse the two narrative migrations" — both already shipped to
  remote and merged; the interim window is in the past.
- "user_id on row vs separate audit log" — architectural debate; tracked
  but out of scope for this PR.
- Multi-signer fastställelseintyg + DB-level period-lock trigger — bigger
  scope, each merits a focused follow-up.

Verification
- 89 unit tests pass
- Zero typecheck errors on any touched file
- Zero lint errors on any touched file
- Migration 20260517180000 applied to remote Supabase via MCP

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-17 15:38:36 +02:00
Jakob Wennberg fd4274787c feat(bokslut): Phases 4-7 — accruals + årsredovisning + EF + MCP tools (#509) 
* feat(bokslut): Phases 4-7 — accruals + årsredovisning + EF + MCP tools

Builds on the Phase 1-3 PR (#508). Same K2-AB scope decisions hold (iXBRL
deferred, K3 deferred, koncernredovisning out of scope) — what ships:

PHASE 4 — Periodiseringar
- lib/bokslut/accruals/accrual-detector.ts: auto-proposes the vacation-
  liability change (delta on 2920 against 7090 + 31,42 % avgifter on
  7519/2940) by reading lib/reports/vacation-liability.ts. Manual prepaid
  (17xx) / accrued (29xx) / audit-fee builders for entries the heuristic
  can't derive — supplier-invoice service_period detection is deferred
  until the data model grows the field.
- /api/bookkeeping/fiscal-periods/[id]/accruals (GET + POST) — posts each
  accrual as a separate manual-source journal entry with the next-day
  reverse date embedded in the description.
- New AccrualsStep wizard step between Preflight and Dispositions.
- Auto-reversal cron is follow-up infra; for now reverses_on is metadata
  + a visible UI badge.

PHASE 5 — Årsredovisning PDF + signing
- lib/bokslut/arsredovisning/{types,build-data,arsredovisning-pdf,signature
  -service}: pre-fills flerårsöversikt from prior 3 fiscal periods, eget-
  kapital-förändring from journal data, K2-minimum noter with
  avskrivningstider auto-derived from the asset register and medelantal
  anställda from the employees table. PDF via @react-pdf/renderer mirroring
  the income-statement/pdf pattern.
- Migration 20260516170000: arsredovisning_signature_requests with RLS,
  signed-immutability trigger, and DELETE policy that blocks signed rows.
  Signature-service exposes list/create/markSigned/isFullySignedOff.
- BankID call itself is not wired here — the table + service make the
  request layer available so a follow-up can hook lib/auth/bankid.ts to the
  sign action without rework.
- /api/.../arsredovisning (data + pdf + signatures) endpoints.
- /bookkeeping/year-end/arsredovisning page with editable narrative,
  flerårsöversikt table, signer slots, PDF download, and Bolagsverket
  Mina Sidor link. Explicit warning about the FY2026 iXBRL mandate.

PHASE 6 — Enskild firma NE-bilaga UI
- lib/bokslut/enskild-firma/{egenavgifter,rantefordelning,periodiseringsfond
  -ef,expansionsfond}-calculator.ts. All declaration-only — never produce
  a journal entry. Egenavgifter 28,97 % / pensionärssats 10,21 % / passive
  SLP 24,26 %, schablonavdrag 25/10/20 %. Räntefördelning SLR+6 / SLR+1
  with the -500 000 negative threshold. P-fond EF cap 30 % (vs 25 % för
  AB). Expansionsfond 125,94 % av kapitalunderlag, 20,6 % skatt.
- EfDeclarationSection mounted inside DispositionsStep when entity_type is
  enskild_firma — live recompute as the user adjusts kapitalunderlag,
  prior-year amounts, p-fond desired, expansionsfond change. Each card
  shows the NE-bilaga ruta the number lands in. NE-bilaga preview link.

PHASE 7 — Agent-native MCP tools
- gnubok_propose_dispositioner: read-only AB dispositions proposal
- gnubok_propose_accruals: read-only accruals proposal
- gnubok_propose_annual_depreciation: read-only depreciation per asset
- gnubok_post_annual_depreciation: stages depreciation commit (high-risk)
- gnubok_preview_arsredovisning: structured K2 ÅR preview
- gnubok_preview_ef_declaration: EF skattemässiga justeringar preview
- All ≤280-char descriptions, additionalProperties:false, conform to
  STAGED_OPERATION_SCHEMA for write tools. Read tools return the same
  shapes as their HTTP counterparts so agents and the UI share a contract.
- The existing GET /bokslutsdispositioner endpoint now calls the shared
  buildDispositionsProposal helper that the MCP tool also uses, removing
  the duplicate logic that lived in both.

Verification
- 149 unit tests pass (was 125 on Phase 1-3; +24 across accruals, EF
  calculators, expansionsfond, periodiseringsfond-ef)
- Zero lint or typecheck errors on any new file
- Migration 20260516170000 applied to remote Supabase via MCP

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bokslut): address PR #509 round-1 — CI blocker + 3 P1s + compliance bundle

CI BLOCKER (output-schema test)
- The 5 new read-only MCP tools (gnubok_propose_dispositioner / _accruals /
  _annual_depreciation, _preview_arsredovisning / _ef_declaration) were
  missing outputSchema, which the existing strict-schemas guard reads as a
  hard failure on core-only CI. Added a permissive { type: 'object',
  additionalProperties: true } outputSchema to each — the return shapes are
  the same as their HTTP counterparts and trying to mirror them inline
  would duplicate the type tree across two boundaries.

P1 — vacation accrual: two real bugs in one entry
- Delta was anchored on the OPENING balance of 2920, so any mid-year
  postings (partial accruals, reversals) were ignored. Now anchors on the
  current closing balance via tb.rows.find(2920).closing_credit -
  closing_debit. Updated the computation field label to current_2920.
- More importantly: 2920 is a balance-sheet carry-forward (semesterlöneskuld
  persists until the actual vacation is paid). The original implementation
  set reverses_on to Jan 1 of the next year, which would zero the liability
  on day 1 of the new year — a known Swedish bookkeeping error. The
  vacation proposal now ships with reverses_on = '' to suppress the
  reversal badge, the API route emits a "Bokslutsjustering" description
  instead of "Periodisering (vänds …)", and AccrualsStep renders "Rullas
  vidare (ingen vändning)" so the user knows the liability carries forward.

P1 — signature_requests immutability gap (security)
- The existing trigger only guarded role / signer_name / signed_at /
  status on signed rows, leaving bankid_signature_data,
  signer_personnummer_encrypted, and signer_personnummer_hash mutable. An
  UPDATE on a signed row could silently alter the BankID proof. New
  migration 20260517090000 replaces the trigger function to cover the full
  audit-critical column set (plus fiscal_period_id and company_id as
  belt-and-braces).

P1 — narrative edits never reached the PDF
- The /bookkeeping/year-end/arsredovisning page let the user edit
  description / important_events / resultatdisposition but the download
  link pointed at a plain GET that regenerated boilerplate. Wired the PDF
  endpoint to accept description / events / disposition as query params
  (length-capped) and the page now constructs the download URL with the
  current narrative state — only fields the user actually changed are
  included, keeping the URL short for the unchanged-defaults case.

Compliance quick wins
- Added period_lock check to gnubok_post_annual_depreciation MCP tool
  (matches the existing accruals POST guard).
- Added explicit fiscal-period ownership pre-check to the signatures POST
  route (RLS would reject anyway; the route layer just makes the 404
  envelope cleaner).
- Replaced free-text role on the signatures schema with an enum allowing
  only Styrelseledamot / Styrelseordförande / VD / Verkställande direktör.
- Added Cache-Control: no-store + Pragma: no-cache to the ÅR PDF response
  so the document (officer names + financials = personal data) isn't
  cached by any intermediary.
- Sanitized period_end in the PDF Content-Disposition header to dodge
  header-injection via stray chars (defensive — period_end is a date, but
  the cost is one regex).
- Softened the iXBRL warning text on the ÅR page: digital filing is
  proposed by Bolagsverket but not yet enacted; PDF is still valid today.

False positives I'm intentionally not chasing on this round
- Greptile P2 `sate` typo in the address city field — the rename would
  touch the type and every consumer; defer.
- Greptile P2 "computation field label" — already addressed as part of
  the P1 vacation fix above.
- Compliance V2.2 "silent skip" in accruals POST — the silent skip is in
  the UI's empty-row filter, not in the server route. Server validation
  already returns 400 via Zod for any invalid item.

Verification
- 154 unit tests pass (was 149; +5 from re-running the MCP strict-schemas
  + output-schema suites that now include the new tools)
- Zero typecheck errors on any touched file
- Zero lint errors on any touched file
- Migration 20260517090000 applied to remote Supabase via MCP

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bokslut): address PR #509 round-2 — 3 real bugs + 5-item polish

3 real bugs from the round-1 Swedish review + Compliance Swarm re-eval:

- Soliditet inflated by obeskattade reserver. build-data.ts filtered equity
  with startsWith('20') || startsWith('21') — but 21xx (periodiseringsfonder,
  överavskrivningar) are partially deferred tax, not eget kapital. Splitting
  them out per K2 / ÅRL. Filter is now startsWith('20') only.

- Resultaträkning omitted bokslutsdispositioner + skatt rows. K2 RR must
  include 88xx (dispositioner) and 89xx (skatt) before "Årets resultat" per
  ÅRL 3:2 — without them, the printed RR doesn't reconcile to BS 2099 and
  the document is non-compliant for any AB that posted bolagsskatt or
  periodiseringsfond. flattenIncomeStatement now splits the financial
  sections on title (Bokslutsdispositioner / Skatter och årets resultat)
  and emits the K2-required intermediate subtotals: "Resultat efter
  finansiella poster" → dispositioner → "Resultat före skatt" → skatt →
  "Årets resultat".

- Accruals POST had no idempotency. Re-running the wizard (or a retried
  POST after a flaky network) would create duplicate accrual entries that
  distort both the balance sheet and trial balance. New
  findExistingAccrualEntry helper queries the period for an existing
  posted entry whose description matches the kind's stable prefix (or for
  manual prepaid/accrued, the user-supplied description). Duplicates land
  in a new `skipped` array in the response with a reference to the
  existing entry id, rather than producing a second posting.

5-item compliance polish:

- signer_name on the signatures schema now has .max(200) per GDPR Art.25.2
  data-minimization. Swedish personal names are well under that — the
  bound is a defense against an unbounded-string injection.

- Audit-fee accrual: 6420 is BAS-specific to lagstadgad revision. Bokslut
  fees for a non-revisionspliktigt bolag (liability_account = 2991) now
  debit 6590 (övriga externa tjänster) instead — Skatteverket may query a
  6420 debit when there is no revisor i bolaget.

- Räntefördelning ne_ruta label: was 'R30 / INK1 kapital' (confusing —
  INK1 is a separate form). Positive now reads 'R30 (avdrag i
  näringsverksamhet)' with the INK1 T4 cross-reference moved to the
  description; negative reads 'R30 (tillägg till resultat)'.

- gnubok_post_annual_depreciation MCP tool now checks the caller's
  company_members role and throws on viewer. RLS would reject the
  underlying INSERT anyway; failing fast here produces a cleaner error
  than the cascaded RLS rejection. Mirrors the HTTP route's
  { requireWrite: true } guard.

- Signature DELETE policy now blocks both 'signed' AND 'declined' rows. A
  declined signature is auditable evidence (board member refused) and is
  material under ABL 8 kap. New migration 20260517100000.

Bot-flagged items I'm deliberately not chasing on this round:
- V8.2.1 × 2 cross-tenant findings on ÅR routes — same false-positive
  class I've responded to repeatedly: buildArsredovisningData internally
  filters by company_id, bot can't see past the route handler.
- V4.5 / V2.2 MCP arg Zod redundancy — MCP server's central handler
  validates against each tool's inputSchema.
- Narrative-in-URL GDPR concerns — proper fix is POST + body or
  server-side persistence; tracked as follow-up.
- `sate` typo (Greptile P2) — type-wide rename, deferred.
- Vacation avgifter age-tier rate split — inherited from upstream
  generateVacationLiability which doesn't expose age; needs upstream
  work.
- Medelantal anställda proper monthly average — needs salary-run
  aggregation across the year, follow-up.
- K2 noter aktiekapital + fastställelseintyg blocks — real K2 gaps,
  tracked as follow-up.

Verification
- 94 tests pass (subset for bokslut + MCP suites; full Phase 1-7 suite
  unchanged in scope)
- Zero typecheck errors on any touched file
- Zero new lint errors on any touched file (the 2 server.ts warnings are
  pre-existing)
- Migration 20260517100000 applied to remote Supabase via MCP

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-17 14:50:48 +02:00
Jakob Wennberg 05078c9d8e feat(bokslut): year-end wizard with bokslutsdispositioner + asset register (#508) 
* feat(bokslut): year-end wizard with bokslutsdispositioner + asset register

Ships the first user-visible bokslut surface for K2 aktiebolag. The year-end
engine, INK2/INK2R/INK2S generator, and reconciliation reports already existed
in lib/core/bookkeeping/ and lib/reports/; this work wires them into a real
multi-step UI, adds the missing dispositioner calculators (bolagsskatt,
periodiseringsfond, överavskrivningar, SLP), and introduces a fixed-asset
register that feeds planenliga avskrivningar into the same flow.

PHASE 1 — Wizard around the existing year-end engine
- Replaces the "Kommer snart" stub at /bookkeeping/year-end with a 4-step
  wizard (Kontroll → Dispositioner → Förhandsgranska → Verkställ) plus a
  Klart result view
- New aggregator lib/bokslut/readiness-aggregator.ts composes
  validateYearEndReadiness with bank-reconciliation status and entity-typed
  reminders into one fetch backing the preflight step
- New endpoint GET /api/bookkeeping/fiscal-periods/[id]/bokslut-readiness

PHASE 2 — Bokslutsdispositioner calculators
- lib/bokslut/tax-provision/{bolagsskatt,sarskild-loneskatt}-calculator.ts —
  20.6 % on taxable result → 8910/2512 (with non-deductible / non-taxable
  manual adjustments and schablonintäkt pass-through) and 24.26 % SLP on
  posted pension costs → 7533/2514
- lib/bokslut/reserves/periodiseringsfond-service.ts — proposeAvsattning
  (25 % cap, BAS 212X cohort accounts) + proposeAteforing (FIFO 6-year
  mandatory reversal with schablonintäkt computation) + balance lookup
- lib/bokslut/reserves/overavskrivningar-service.ts — 30-rule + 20-rule
  helpers + proposeOveravskrivningar (8853/2153)
- New endpoint /api/bookkeeping/fiscal-periods/[id]/bokslutsdispositioner
  (GET ordered proposals, POST commits user-chosen ones as separate
  year_end vouchers via the journal engine)
- New DispositionsStep UI: per-card accept/skip + editable amount where
  meaningful; mandatory p-fond reversals can't be skipped
- INK2 bug fix: ink2-engine.ts SRU mapping ranges previously pointed at
  accounts BAS doesn't seed (8810/8830/8840). Corrected to 8811 (avsättning),
  8819 (återföring), 8830 (lämnade koncernbidrag) so calculator output now
  flows into INK2 correctly. Regression-locked with 6 new mapping tests.

PHASE 3 — Anläggningsregister + depreciation engine
- New migration 20260516120000_assets_and_depreciation.sql: assets table
  (category, BAS-triple, K3 components JSONB reserved) and
  depreciation_schedules (asset+period+journal_entry link). RLS via
  user_company_ids(), immutability triggers after disposal/posting.
- lib/bokslut/assets/asset-service.ts — CRUD + disposal that posts a proper
  gain/loss entry against 3973/7973
- lib/bokslut/assets/depreciation-engine.ts — computeAnnualDepreciation
  (linear, pro-rata at acquisition/disposal/end-of-life) +
  proposeAnnualPostings + commitAnnualPostings (one entry per asset)
- New endpoints /api/assets (CRUD + dispose) and
  /api/bookkeeping/fiscal-periods/[id]/depreciation (preview + commit)
- /assets list+create page with K2 schablon defaults (3y datorer,
  5y inventarier, 25y byggnader); sidebar entry added
- DepreciationPanel mounted at the top of DispositionsStep; posting
  refreshes dispositions so bolagsskatt picks up the new result

Out of scope (per the agreed plan): K3 framework, iXBRL filing to
Bolagsverket (manual export only for now — regulatory risk flagged for
FY2026 closings), inventory module, koncernredovisning, revisor workflow.

Verification
- 116 unit tests pass across lib/bokslut/, lib/reports/ink2/, and the
  existing lib/core/bookkeeping/year-end-service suite
- Zero lint or typecheck errors in any new file
- Migration applied successfully via Supabase MCP

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bokslut): address PR #508 review — P1 correctness + P2 conventions

P1 — stale reminders in readiness-aggregator (greptile)
Remove the depreciation_manual / bolagsskatt_manual / periodiseringsfond_manual
nudges. After Phase 3, the wizard handles all three automatically via
DepreciationPanel and the dispositioner calculators — surfacing them as
manual reminders told users to repeat work the page already did. Only the
accruals_manual reminder (Phase 4 hasn't shipped) and the EF-only
ef_skatt_via_ne reminder remain.

P1 — netBookValueAfter ignored prior accumulated depreciation
proposeAnnualPostings now fetches all prior posted depreciation_schedules
for the company (excluding the current period) and sums them per asset, so
the displayed restvärde reflects every previously-booked year of avskrivning
instead of only this year. Without the fix, a 5-year asset in year 3 would
have shown 48 000 instead of the correct 24 000 net book value.

P1 — ordering bug in dispositioner POST handler
The 25 % p-fond avsättning cap derives from the current trial balance, so
mandatory återföring entries must post first. Added a server-side sort by
canonical bokslut order (återföring → överavskrivningar → avsättning → SLP
→ bolagsskatt) regardless of the client array order. The cap can no longer
be evaluated against a stale pre-återföring net result.

P2 — depreciation_schedules missing updated_at
New migration 20260516140000_depreciation_schedules_updated_at.sql adds the
column + trigger via update_updated_at_column(). Per CLAUDE.md migration
conventions, never modified the original migration. DepreciationSchedule
type updated.

P2 — addMonths end-of-month overflow
Replaced setUTCMonth (which overflows: Jan 31 + 1 month → Mar 3) with a
day-clamping implementation that produces Feb 28/29. Without the fix,
lifeEndExclusive landed one day too late and slightly over-depreciated.
New regression test asserts Jan 31 + 12 months stays in January.

P2 — pg-real tests for new triggers and RLS
tests/pg/assets.pg.test.ts (13 tests) covers:
  - enforce_asset_post_disposal_immutability blocks every financial field
    after disposal, allows notes/name through
  - assets_disposal_atomic CHECK requires both disposed columns set together
  - enforce_depreciation_schedule_immutability blocks edits after
    journal_entry_id is linked, allows them before
  - depreciation_schedules delete RLS policy filters out posted rows
  - assets + depreciation_schedules RLS isolates across companies

Verification
- 117 unit tests pass (was 116, +1 for the addMonths regression)
- New pg-real suite syntactically + type-correct; will execute in CI
- Zero lint or typecheck errors in any touched file
- Migration 20260516140000 applied to remote Supabase via MCP

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bokslut): address compliance-swarm + swedish-review findings

Real bugs surfaced by the bots on the first push that weren't covered by
greptile's inline P1/P2 set:

- Immaterial asset disposal hit the wrong BAS accounts. disposeAsset always
  posted gain/loss to 3973/7973 regardless of category. For category
  'immaterial' it must use 3013 (vinst) / 7813 (förlust) per BAS — using the
  tangible accounts misclassifies in INK2R. Now branches on category. Two
  new regression tests pin each branch.

- acquisition_cost CHECK was too loose. CreateAssetSchema accepted 0 (just
  nonnegative). Tightened to z.number().positive() — a zero-cost asset
  creates a no-op depreciation row and a balance sheet line that nothing
  reconciles against.

- UpdateAssetSchema let users remap BAS accounts arbitrarily. Bot flagged
  this as a defense-in-depth gap (V4.5). Added BAS_RANGES_BY_CATEGORY
  validation at both the schema layer (Create) and the service layer
  (Update) so user-supplied account overrides must stay inside the
  category's expected BAS range. INK2R mappings and the depreciation
  engine's category-driven defaults now can't drift.

Swedish accounting review:

- Building/markanläggning defaults — clarified UI copy. The 25-year
  schablon is K2-redovisning, not the IL skattemässig rate. New helper text
  spells this out. Markanläggning default lowered from 20→10 years
  (Skatteverket guidance allows 10 % rate; 20 was on the upper bound
  without justification).

- createAsset doesn't post the acquisition entry by design — that gap
  wasn't called out anywhere in the UI. Added a tip box in
  CreateAssetDialog explaining that the acquisition must already be in the
  books; the register only drives depreciation.

- Disposal VAT (ML 3:3 / 7:3) not handled — sale of a deduct-eligible
  anläggningstillgång is in principle 25 % momspliktig. Documented this as
  a known limitation in the disposeAsset docstring so any future UI
  surfacing the disposal endpoint warns the user.

Documented (not fixed yet) — bot was right but wider-scope work:

- SOC 2 PI1.3: dispositioner POST loop is not transactional. A failure
  midway leaves partial postings. Added a code comment explaining the
  recovery path (re-POST omitting committed kinds — each calculator
  re-derives from current TB). Real atomicity via an RPC wrapper is Phase
  5+ work.

False positives intentionally not changed:

- 4× OWASP V8.2.1 cross-tenant findings — service functions already filter
  by company_id; the bot can't see past the route handler.
- V2.3 client-supplied amount clamping — proposeAvsattning and
  proposeAteforing both clamp via Math.min already.
- A.8.15 audit events — withRouteContext already logs completion.
- Schablonintäkt journal entry — per IL 30:6a it's a skattemässig
  justering, never booked. Current implementation is correct.
- Voucher series 'A' — matches existing executeYearEndClosing convention;
  not changing here in isolation.

Verification
- 119 tests pass (was 117, +2 for the immaterial-disposal branches)
- Zero lint or typecheck errors on any touched file

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bokslut): address compliance round 4 — BAS account overlap + method gate

Two real bugs the compliance bot caught after my round-3 fixes (both new on
its re-run, not in the original report):

- BAS account overlap not prevented: my BAS_RANGES_BY_CATEGORY uses the same
  class range for asset and accumulated (e.g. immaterial: both 1010–1099,
  building: both 1100–1199). Nothing stopped a user from picking the same
  account for both, which would silently net acquisition cost against
  accumulated depreciation in one bucket and corrupt INK2R 720x mappings.
  CreateAssetSchema now rejects bas_asset_account === bas_accumulated_account
  in a superRefine cross-field check; updateAsset enforces the same invariant
  by reading the existing asset and validating the merged result.

- declining-balance methods silently fell back to linear. The DB enum allowed
  declining_balance_30 / declining_balance_20, but the engine's
  computeAnnualDepreciation only implements linear math. A determined caller
  (MCP, curl, future UI) could create an asset labelled as räkenskapsenlig
  avskrivning and get linear charges — silently wrong numbers under a
  misleading method. Both CreateAssetSchema and UpdateAssetSchema now refine
  the depreciation_method enum to require 'linear'. The DB enum stays open
  for a future phase to add proper support. Stale comment in
  depreciation-engine.ts updated to reflect the new invariant.

False positives I'm explicitly not chasing further on this round:
- 3× repeated OWASP V8.2.1 cross-tenant — services already filter by
  company_id; bot can't see past the route handler. Round 3 already added
  service-layer tests and inline reasoning.
- V2.3 atomicity upgrade to high — bot now flags it harder *because* I
  documented it in round 3. The existing executeYearEndClosing has the same
  non-transactional sequential-write pattern; wrapping just this endpoint
  in an RPC while leaving the rest inconsistent is worse than the doc
  comment. Real atomicity is Phase 5+.
- Disposal VAT user-facing warning — no UI surfaces dispose yet; docstring
  in the service is sufficient until the UI ships.

Verification
- 119 tests pass
- Zero lint or typecheck errors on any touched file

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bokslut): address compliance round 5 — disposal integrity + öavskr accounts

Round-5 bot feedback after my round-4 push (the sticky reports re-edited
themselves with two new genuine findings; bot count: 18 → 14 active).

P1 — accumulated_depreciation was client-supplied (OWASP V8.2.1, swedish
compliance review): the dispose endpoint accepted accumulated_depreciation
as a request-body number. A malicious or buggy caller could inflate it to
manipulate the book-value calculation and pocket a phantom gain. Now:

- DisposeAssetSchema no longer accepts accumulated_depreciation
- disposeAsset sums planned_depreciation from depreciation_schedules where
  journal_entry_id IS NOT NULL for the asset, server-side
- New regression test "server-derives accumulated_depreciation — caller
  cannot inflate gain" pins the server-derivation against the prior attack
- Limitation: manual avskrivningsverifikationer posted outside the engine
  aren't captured. Phase 5+ can swap this for a trial-balance scan on
  bas_accumulated_account if that gap matters.

P2 — överavskrivningar hardcoded 8853/2153 regardless of asset category
(swedish-asset-accounting): for buildings BAS uses 8852/2152 and for
immateriella tillgångar 8851/2151. Edge case for K2 SME (öavskr on
buildings is rare; on immateriella rarer still) but worth not lying about
the accounts. Now:

- New OVERAVSKRIVNING_ACCOUNTS table maps category → expense/accumulated
  pair (machinery_equipment, building, immaterial, group)
- proposeOveravskrivningar accepts optional category, defaults to
  machinery_equipment (the dominant K2 case — no behaviour change for
  existing callers)
- POST handler item schema accepts optional category
- Label + description strings now name the actual accounts used
- 3 new tests cover the building, immaterial, and default branches

False positives I'm still declining to chase (already covered in prior
commit messages):
- 3× repeated OWASP V8.2.1 cross-tenant — services scope by company_id;
  bot can't see past route handler
- V2.3 atomicity — existing executeYearEndClosing has the same pattern;
  wrapping just this endpoint is inconsistent; real fix is Phase 5+ RPC
- Disposal VAT user-facing warning — no UI surfaces dispose yet

Verification
- 123 tests pass (was 119, +3 for öavskr category branches and +1 for the
  server-derivation regression test, with one prior test rewritten to use
  the new server-supplied accumulated path)
- Zero lint or typecheck errors on any touched file

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(bokslut): pre-merge polish — building disposal accounts, SLR-2026 rate

Last polish round before merge. All three from the round-5 sticky
Swedish-accounting-review update:

- Building / markanläggning disposal posted gain/loss to 3973/7973. BAS
  2026 routes those to 3971/7971 (the SRU mapping points them at a
  different INK2R field, so the existing accounts misclassify). Extended
  the existing immaterial branch (3013/7813) into a three-way:
    immaterial            → 3013 / 7813
    building / land_imprv → 3971 / 7971
    other tangible        → 3973 / 7973
  Two new regression tests pin the building and land_improvement branches.

- DEFAULT_SCHABLONINTAKT_RATE was 0.03, based on SLR 2024-11-30 (1.96 %).
  For closings of inkomstår 2026 the rate is SLR 2025-11-30 (2.55 %) + 1 pe
  = 3.55 %. The wrong rate under-taxes the schablonintäkt, which feeds into
  bolagsskatt. Updated to 0.0355 and rewrote the doc comment to track both
  years so the next bump is obvious.

- Jämkning of input VAT for buildings / markanläggning disposed within the
  10-year jämkningsperiod (ML 9 kap 8–11 §§) is out of scope for this PR
  but should not be silently absent — added a KNOWN LIMITATION block to the
  disposeAsset docstring so any future UI surfacing disposal checks the
  10-year window and warns the user.

After this push the PR has 125 passing tests, all CI green, no merge
conflicts, and the only remaining bot complaints are repeat false
positives or Phase 5+ scope (RPC atomicity, full asset disposal UI,
K3 component depreciation).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-16 15:55:27 +02:00
Jakob Wennberg c06395f633 feat(mcp): agent-native API sprint — quick wins (items 8/10/38/39/50) (#505) 
* feat(mcp): agent-native API sprint — quick wins (items 8/10/38/39/50)

Five Tier-S items from dev_docs/api_ai_architecture/PLAN.md, picked for highest
impact-per-day on a solo budget. ~7.5 engineer-days of work.

Item 38 — gnubok_reverse_journal_entry MCP tool. Wraps the existing
reverseEntry() engine function (lib/bookkeeping/engine.ts) as a staged
high-risk operation. Description distinguishes pure makulering (use this) from
rättelse (use gnubok_correct_entry) per BFL 5 kap 5§ guidance — leaving a real
affärshändelse unbooked is itself a BFL violation, so agents must understand
which storno pattern to apply. New operation_type 'reverse_entry' wired through
PendingOperationType, risk-tiers (high), commit.ts executor, and TOOL_SCOPE_MAP
(bookkeeping:write). Six executor cases + three staging-gate cases cover the
new tool.

Item 39 — period_status threading. New helper resolvePeriodStatusForDate() in
lib/core/bookkeeping/period-service.ts returns { period_id, status, lock_date }
using the same two-layer logic as the v1 REST check (company-wide
bookkeeping_locked_through + fiscal_period flags). Threaded through
stagePendingOperation via a new dateForPeriodCheck option so agents and widgets
can detect locked/closed periods without round-trips. Applied to seven
bookkeeping-touching tools: categorize_transaction, create_transactions,
create_voucher, approve_supplier_invoice, mark_invoice_as_paid, correct_entry,
reverse_journal_entry. Resolution failure is non-fatal — DB triggers stay
authoritative.

Item 50 — gnubok://company/current expansion. Replaces the metadata-only
resource with per-company working memory: active fiscal period status, lock
dates, counts (customers, suppliers, open AR/AP, uncategorized transactions),
voucher series state across open periods, recency signals (last categorization,
last invoice sent, last bank sync), and the next five approaching deadlines.
All queries parallelized via Promise.all; payload stays well under 8 KB.
Mirrors the context.md pattern from Shipper+Claude's agent-native architecture
guidance and prevents the context-starvation anti-pattern.

Item 8 — schema strictness. additionalProperties: false on every one of the 67
inputSchemas in extensions/general/mcp-server/server.ts. New
strict-schemas.test.ts guards against regression on newly authored tools.
CLAUDE.md documents the tool-authoring contract (strict input schemas,
description ≤280 chars, STAGED_OPERATION_SCHEMA + next as the
completion-signal pattern — do NOT introduce a parallel S/H/C/O envelope).
Payload-size ceiling raised from 20K → 25K tokens with a comment pointing at
item 15 (Tool Search + defer_loading) as the long-term answer rather than
relaxing the watchdog further.

Item 10 — prompt cache groundwork. The only Anthropic SDK call site in the
codebase is the invoice-inbox extension's Bedrock-backed extractor; tagged the
~3.5 KB SYSTEM_PROMPT with cache_control: { type: 'ephemeral' } and added
usage logging (cache_read_input_tokens / cache_creation_input_tokens) so the
hit ratio is measurable. The plan's 1h TTL is direct-Anthropic-only;
documented the constraint and the MCP-side determinism contract (tool
definitions must be byte-stable across requests) in the new mcp-server
README.md.

Carry-over: includes a small untracked migration
(20260516060000_journal_entries_source_type_inbox_item) and its pg test guard
that fix a production CHECK-constraint gap for source_type='inbox_item' —
unrelated to the sprint but bundled per request.

Tests: 3615/3615 pass across 252 files. TypeScript build clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): address PR #505 review — cross-tenant leaks, company-wide lock, PII

Five reviewer findings on PR #505 addressed:

1. Cross-tenant leak — voucher_sequences (OWASP V8.2.1, SOC 2 CC6.3).
   Resource query filtered by user_id only; switched to company_id since the
   table has both (added in the 2026-03 multi-tenant refactor migration).

2. Cross-tenant leak — deadlines (OWASP V8.2.1, GDPR Art.5(1)(f), ISO A.8.3).
   Same fix; the deadlines table also gained a company_id column in the
   multi-tenant refactor and the RLS policies enforce it. With the company_id
   filter active, the userId parameter is no longer needed in the resource —
   removed from the destructure.

3. Compliance gap — commitReverseEntry and commitCorrectEntry only checked
   fiscal_periods.is_closed, not company_settings.bookkeeping_locked_through.
   Agents could stage a reversal with period_status: locked warning (caught
   by resolvePeriodStatusForDate at staging time), have the user approve,
   and the commit would slip through. Both executors now run
   resolvePeriodStatusForDate at commit time so the gate matches the
   staging-time signal. Pre-existing gap on commitCorrectEntry also fixed.

4. Schema mismatch — period_status was spread into both `preview` and the
   top-level response, but STAGED_OPERATION_SCHEMA only declares it at the
   top level. Removed the preview-nested copy to match the schema and avoid
   ambiguous reads.

5. Tool description — swedish-compliance bot flagged that "pure makulering
   (storno)" conflates two distinct Swedish accounting terms: storno
   preserves the original; makulering voids it entirely. Code does storno;
   description now says so plainly and cites BFL 5 kap.

6. Input hardening — added ^\d{4}-\d{2}-\d{2}$ pattern to reversal_date in
   inputSchema plus a runtime regex check in execute(), so a malformed date
   never reaches the pending_operations payload.

7. GDPR — ai_extraction_usage and the two pre-existing fileName log
   emissions in extract-invoice-fields.ts replaced raw fileName with a
   12-char SHA-256 prefix. Raw invoice file names (e.g.
   "faktura_Sven_Andersson.pdf") can constitute personal data; hashing
   preserves operator correlation without exposing PII to log destinations
   that may lack documented retention controls.

Notes on findings NOT addressed:
- Double-reversal guard (Greptile/swedish-compliance): false positive.
  reverseEntry() flips the original's status to 'reversed' (engine.ts:538)
  and the staging tool already rejects anything not 'posted'. Engine also
  has a CAS guard at lines 541-551.
- Staging vs commit TOCTOU re-validation: pre-flight + DB triggers remain
  authoritative; the window is narrow enough that adding executor-side
  re-checks isn't load-bearing this sprint.
- Runtime Zod validation of args inside execute(): codebase doesn't do
  this for any MCP tool today; cross-cutting refactor deferred.

New test: voucher-executors.test.ts adds a case for the company-wide lock
branch on reverse_entry (verifies the new resolvePeriodStatusForDate gate
fires when bookkeeping_locked_through covers entry_date).

Tests: 3616/3616 pass. TypeScript build clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): address second-round PR #505 review — locked_at, reason cap, log

Re-review by compliance-swarm and swedish-accounting-compliance bots after the
first fixes raised three more legitimate findings:

1. Per-period `locked_at` not directly checked from the fetched row
   (swedish-accounting-compliance). Both commitCorrectEntry and
   commitReverseEntry already call resolvePeriodStatusForDate which covers
   locked_at, but a transient DB blip in the resolve helper would silently
   skip that gate. Now reading locked_at directly from the inner-join row and
   checking it alongside is_closed before the resolve helper runs — same
   pattern, two defense-in-depth layers instead of one.

2. `reason` field had no maxLength (OWASP V4.5). Added maxLength: 500 to the
   inputSchema and a runtime length check; an adversarial agent could
   otherwise push an arbitrarily large string into pending_operations.

3. periodStatus resolution failure was silently swallowed (ISO 27001 A.8.15).
   Now logging via console.warn with operationType, companyId,
   dateForPeriodCheck, and error so a systematic outage (missing
   company_settings row, dropped query) is observable in audit logs rather
   than degraded silently.

Findings deliberately NOT addressed (pushed back to the bots):

- gnubok_reverse_journal_entry needs per-operation role check (V8.2.1) and
  narrower 'bookkeeping:reverse' scope (CC6.3) — cross-cutting refactor; no
  MCP tool in gnubok enforces per-operation roles today. Introducing it just
  for one tool would be inconsistent. Will surface as a separate item.
- Reduce line_description in reverse_entry preview (A.8.3, Art.5(1)(c)) —
  the preview is shown to the human approver who needs to see what they're
  approving under BFL 5 kap. Aggregate-only previews would harm the
  approval workflow.
- Audit company-current fields for PII (A.8.12, Art.25(1)) — vat_number,
  org_number, etc. are intentionally part of working memory; agents need
  them to make compliant booking decisions.
- Payload-size ADR reference (A.8.9) — the test comment already cites plan
  item 15 (Tool Search) as the long-term answer.
- mime_type classification label (CC7.2) — theoretical concern;
  ai_extraction_usage events are already operator-only.
- False positive: commitReverseEntry already has the closed-period check
  (V2.3); bot was hallucinating.

Tests: 3616/3616 pass. TypeScript build clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp,env): structured logger + description trim + env alias support

Two further follow-ups on PR #505:

1. resolvePeriodStatusForDate catch now uses the structured logger
   (createLogger from @/lib/logger) instead of console.warn. Three
   reviewers (compliance-swarm V16.1.1, ISO 27001 A.8.15, SOC 2 CC7.2)
   independently flagged that console.warn bypasses the centralized log
   aggregation pipeline used elsewhere, so systemic outages of the
   period-status resolver were invisible to the SIEM. log.warn now routes
   through the same sink as other server events.

2. Tool description for gnubok_reverse_journal_entry now routes the refund
   case explicitly to gnubok_credit_invoice. The Swedish accounting
   compliance bot flagged that the previous "cancelled credit invoice"
   example was ambiguous — a real credit invoice flow goes through
   gnubok_credit_invoice, not this tool. Description stays under 280 chars.

3. lib/init.ts: REQUIRED_EXTENSION_VARS now models each entry as a list of
   acceptable aliases instead of a single required name. The fallback in
   extensions/general/enable-banking/lib/jwt.ts already accepts the
   _PRODUCTION-suffixed variants (used by Vercel prod) as equivalent to
   the base names, but the env validator at boot didn't, so every cold
   start in prod warned about missing ENABLE_BANKING_APP_ID even though
   ENABLE_BANKING_APP_ID_PRODUCTION was set and the runtime was healthy.
   Each entry now satisfies if ANY listed alias is present; missing
   entries print all acceptable names so operators can pick either form.

Tests: 3616/3616 pass. TypeScript build clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): staging tools reject locked_at periods too, not just is_closed

Swedish accounting compliance bot flagged that gnubok_reverse_journal_entry
and gnubok_correct_entry pre-flight checks only rejected closed periods —
locked-but-not-closed periods passed staging and only got rejected at
commit time. The commit-time gate was correct (both executors check
is_closed AND locked_at AND resolvePeriodStatusForDate), but the
staging-time signal was confusing: agent saw staged:true with
period_status:"locked" in the same envelope.

Now the staging pre-flight reads locked_at from the same inner-join and
rejects on either flag, matching the commit-time pattern. The error
message updated to "locked or closed" since both branches reach the same
throw. BFL 5 kap 5§ alignment is unchanged — both paths still block
mutations to locked/closed periods; only the layer at which the rejection
fires changes.

Findings pushed back (response in PR thread, not addressed here):
- companyId/mimeType in log.warn flagged as PII (overreach; tenant IDs
  are operational identifiers, not personal data, and the codebase logs
  them consistently elsewhere).
- HMAC-keyed file_name_hash instead of plain SHA-256 prefix (overreach;
  48 bits already addresses the immediate GDPR Art. 5(1)(f) concern).
- 'title' field in deadlines may contain PII (overreach; would require
  redacting every text field in every read resource).
- RLS regression test for voucher_sequences/deadlines (legitimate but
  pg-test scope; tracked for a follow-up sprint).
- Payload-size ADR record (comment already cites plan item 15).
- company-current data minimisation (already pushed back; agents need
  the fields for compliant booking decisions).
- Error message conflates "locked" and "closed" — minor UX nit not
  worth distinguishing here since the remediation step (unlock / omprövning)
  is the same for the user.
- reversal_date period attribution & voucher series integrity flagged as
  unverifiable from diff — false positives, both already handled by the
  engine (period_id from original, atomic voucher number).

Tests: 3616/3616 pass. TypeScript build clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(mcp): address Swedish-accounting compliance round 4 — BFL invariant + VAT warning

Three legitimate findings from the swedish-accounting-compliance bot acted on
(out of five total; two pushed back as theoretical/false positive):

1. BFL 5 kap 5§ invariant assertion (finding 1). The engine guarantees that
   reverseEntry() posts the storno to original.fiscal_period_id (engine.ts:492
   — verified by reading the code), but the executor previously took that on
   faith. commitReverseEntry now asserts reversal.fiscal_period_id ===
   original.fiscal_period_id after the call and returns a 500 with an
   explicit "BFL invariant broken" error if the engine ever drifts. New
   executor test covers this. The reversal_date parameter is unchanged —
   it's used as the storno's entry_date (operational date), not for period
   attribution, per BFL practice (entry_date can differ from period_id's
   range for a rättelse made later).

2. resolvePeriodStatusForDate unhandled-rejection path (finding 2). Both
   commitCorrectEntry and commitReverseEntry now wrap the resolve call in
   try/catch, returning a clean Swedish 500 instead of letting the
   dispatcher surface a raw Postgres error message. Matches the
   log-and-degrade pattern already used at staging time in
   stagePendingOperation.

3. VAT-period warning in the reverse preview (finding 4 — swedish-vat).
   When the original entry contains 2610–2670 BAS accounts, the staged
   preview now includes a Swedish warnings[] field telling the approver
   that a storno is legally insufficient if the moms period has been
   filed with Skatteverket — they must use omprövning per ML 2023:200
   instead. Soft warning (not a hard block) since gnubok doesn't track
   per-VAT-period filing status today; the human decides at approval.

Pushed back:

- Finding 3 (TOCTOU between staging and commit on fiscal_period_id):
  posted entries are immutable per the enforce_journal_entry_immutability
  trigger (migration 20240101000017). fiscal_period_id can't change
  between staging and commit. Status change is already caught by the
  status !== 'posted' check.

- Finding 5 (migration 20260516060000 not wrapped in BEGIN/COMMIT):
  Supabase migration tooling runs each migration file in an implicit
  transaction. PostgreSQL DDL is transactional. The DROP/ADD pair is
  atomic in practice. The bot acknowledges this as low severity.

Tests: 3617/3617 pass (one new — BFL invariant assertion). Build clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-16 11:42:47 +02:00
Jakob Wennberg afb21ea638 feat(api): Phase 6 PR-3 — substrate hardening (SKIP LOCKED + DNS pinning + test debt) (#500) 
* feat(api): operations table immutability trigger

BFNAR 2013:2 kap 8 § behandlingshistorik integrity: once an operations
row is in a terminal status (succeeded / failed / cancelled) the audit
record of what happened becomes immutable. Adds the BEFORE UPDATE and
BEFORE DELETE triggers that the webhook_deliveries table already has
(20260515170000 / 20260515190000), mirroring their predicate shape and
error code exactly.

Closes the Phase 4 PR-2 (PR #469) review-round carry-over flagged by
Swedish-compliance: previously a future bug, a privileged operator, or
a compromised service-role caller could rewrite "this year-end close
succeeded" to "failed" by updating an already-terminal row. The
running → succeeded/failed/cancelled transition itself stays legal
because the trigger keys on OLD.status, which is non-terminal at the
moment of the legitimate UPDATE.

pg test covers all transitions (allowed and blocked) plus DELETE on
both terminal and non-terminal rows.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(api): atomic SKIP LOCKED claim for webhook dispatch

Replaces the SELECT-then-UPDATE-intersect pattern in the dispatcher
with a single-roundtrip SQL function using FOR UPDATE SKIP LOCKED.
PostgREST can't express SKIP LOCKED through the JS client, so the
previous shape relied on a CAS guard inside an UPDATE WHERE status IN
('pending','failed') to ensure only one of two overlapping cron ticks
claimed any given row.

The CAS pattern was correct (under load — receivers >60s could push a
batch past the next minute's tick) but burned two round trips and
forced the application to negotiate the locking semantics in JS.
The function form moves the contention to the DB, where SKIP LOCKED
makes a row held by a concurrent tick simply invisible to the second
caller. One round trip, no JS-side intersect.

All filter semantics are preserved verbatim inside the function:
status IN ('pending','failed'), next_attempt_at <= now, webhook_id
IS NOT NULL, ORDER BY next_attempt_at ASC, LIMIT batchSize. p_batch_size
is bounded (0, 1000] to forestall a runaway lock-set in case a caller
misconfigures it.

pg test covers basic claim (pending + failed), future-due skip, dangling-
row (webhook_id IS NULL) skip, terminal-status skip, batch-size limits,
out-of-range argument rejection, and the SKIP LOCKED invariant itself
using two concurrent pool clients in BEGIN — the second caller does not
see the row A locked, no double-delivery.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(api): pinned-IP HTTPS dispatch (close DNS-rebinding window)

The url-guard.ts file header openly flagged the remaining gap:
"a separate DNS-rebinding window (between dispatch-time validation
and the actual fetch) remains; closing that requires a custom HTTPS
agent that pins the resolved IP — tracked for follow-up." This closes it.

The previous shape was:
  1. validateWebhookUrl()  → DNS resolves to [public IP], returns ok
  2. fetch(webhook_url)    → re-resolves DNS; an attacker who flipped
                             the A record in the interval gets a
                             private-IP socket

The new pinnedHttpsFetch helper validates DNS once, then opens a
node:https.request to that pinned IP — but keeps the original hostname
in the TLS SNI extension (so the receiver's cert validates) and in the
HTTP Host header (so vhost routing still works). The request socket
never re-resolves DNS, foreclosing the rebind race entirely.

Built on node:https.request rather than undici's Agent so the project
doesn't take on a new dep — the stdlib API is also more explicit about
the SNI / Host / pinned-IP split. Test seam injects both validateUrl
and httpsRequest so the unit tests verify the pinning shape without
standing up an HTTPS server.

The dispatcher's attemptDelivery is rewritten as a switch over the four
PinnedFetchResult kinds (ok / unsafe_url / redirect_blocked / timeout
/ transport_error). The previous fetch-based code path that distinguished
redirect rejection by string-matching err.message is gone — the new
result type makes the distinction structural.

8 unit tests cover the SNI/Host/pinned-IP shape, port handling,
redirect_blocked, transport_error, timeout, response-body truncation,
first-IP determinism, and the validation short-circuit (never opens a
socket when the URL fails the SSRF guard).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* test(api): pg tests for webhook substrate triggers (PR-1 test debt)

CLAUDE.md ("Testing" + "Migration Rules") mandates a *.pg.test.ts for
any PR touching a trigger / RPC / RLS / DEFERRABLE constraint. Phase 6
PR-1 (#496) shipped three webhook_deliveries triggers without the
accompanying pg test; this closes that debt.

Triggers covered:
  - enforce_webhook_delivery_immutability  (BEFORE UPDATE)
  - block_webhook_delivery_terminal_delete (BEFORE DELETE)
  - assert_webhook_delivery_company_match  (BEFORE INSERT)

13 cases verify the lifecycle the dispatcher depends on remains mutable
(pending → in_flight, in_flight → failed, failed → in_flight, in_flight
→ delivered) while terminal-status rows (delivered / dead) are write-
locked and the cross-tenant INSERT path is refused with the
ERRCODE=check_violation contract documented in the migration.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* test(api): integration tests for webhook routes (PR-1 test debt)

CLAUDE.md mandates integration tests under app/api/v1/ for every route.
Phase 6 PR-1 (#496) shipped the eight v1 webhook routes (five under
/companies/{companyId}/webhooks/ + the cross-tenant /webhook-deliveries/
{id}/retry) without them; closes that debt.

19 cases for the /webhooks/ verticals:
  POST     /webhooks                    create + secret-once + payroll-scope gate + SSRF
  GET      /webhooks                    list (no secret) + empty list
  GET      /webhooks/:id                detail (no secret) + 404
  PATCH    /webhooks/:id                update + active=true re-enable + SSRF re-check + empty-body
  DELETE   /webhooks/:id                204 hard delete
  POST     /webhooks/:id/test           enqueue + 404 + disabled-rejection
  GET      /webhooks/:id/deliveries     happy path + ownership 404

7 cases for the retry route:
  POST /webhook-deliveries/:id/retry   dead → fresh pending row, live-status refusal,
                                       cross-tenant 404, disabled-webhook gate,
                                       SSRF re-check, delivery 404, webhook-gone 404

Both files mirror the suppliers/customers integration test pattern:
Proxy-backed Supabase mock with per-table queues, validateApiKey +
validateWebhookUrl stubbed to control auth and DNS deterministically.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* refactor(api): address PR-500 review round 1 — pg-real CI fix + 4 review items

1. pg-real CI was red on this PR: the new webhook trigger pg.test.ts and
   claim-due-webhook-deliveries pg.test.ts fixtures tried to INSERT into
   `webhooks.user_id`, which doesn't exist in the migration history. The
   column was never declared in automation_webhooks (20260415000000) nor
   added by webhooks_v2 (20260515170000) — so a fresh schema replay had
   no such column. The webhook create route (`webhooks.create`) was also
   referencing this non-existent column in its INSERT, so the production
   route was latent-broken since PR-1 and never exercised against a fresh
   DB. Drop the `user_id` field from both the route INSERT and the pg
   fixtures. Actor attribution lives on `created_by_api_key_id` (which
   leads back to the owning user via `api_keys.user_id`).

2. Greptile P2 #1 — `recoverStuckInFlight` carried a redundant
   `.not('status','in','(delivered,dead)')` filter alongside
   `.eq('status','in_flight')`, with a comment that incorrectly described
   PostgreSQL's UPDATE re-evaluation semantics. Under READ COMMITTED,
   UPDATE re-evaluates WHERE against each row's CURRENT value when it
   acquires the row lock — a row that raced to terminal status will fail
   `status='in_flight'` on re-evaluation and be skipped, no immutability
   trigger fires. Drop the redundant filter and rewrite the comment.

3. Greptile P2 #2 — added explicit pg test verifying `in_flight` rows are
   skipped by `claim_due_webhook_deliveries`. The status filter is what
   prevents double-delivery and is the entire point of the SKIP LOCKED
   substrate; making that invariant load-bearing in the test suite
   forecloses a future filter expansion silently regressing it.

4. Greptile P2 #3 — pinned-fetch registered both `res.on('end', finalize)`
   and `res.on('close', finalize)`. Node fires BOTH on normal completions,
   so finalize ran twice; the outer `settled` guard squashed the
   double-resolve but the header reconstruction still ran twice. Switch
   to `once` + self-removing pair so finalize runs exactly once on
   whichever event fires first (normal: end; truncation: close).

5. Compliance Swarm V8.2.1 — the retry route only checked
   `webhooks:manage` even when retrying `salary_run.* / agi.*` deliveries.
   Mirror the create-route elevated-scope gate so a key with only
   `webhooks:manage` cannot re-emit payroll payloads carrying
   personnummer / lönesummor / skatteavdrag. New integration test verifies
   the gate returns 403 INSUFFICIENT_SCOPE with `required_scope:
   payroll:read`.

35 tests pass locally (+1 vs pre-fix). Type-check clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* refactor(api): address PR-500 review round 2 — 2 small precision fixes

1. Compliance Swarm Art.32 / A.8.24 — response_body size cap was enforced
   only at the application layer (pinnedHttpsFetch's maxResponseBytes=4096
   constant). A future refactor that bypassed the truncation, or a non-
   dispatcher write path into webhook_deliveries.response_body, would
   silently land large blobs in a column adjacent to event payloads
   carrying personal data. Add a CHECK constraint at the DB layer with
   a generous ceiling (8 KB — double the application cap so legitimate
   dispatcher writes never hit it; only a regression surfaces as a
   check_violation).

2. Compliance Swarm CC6.6 — pinned-fetch substitutes the validated IP
   for `host` while keeping the original hostname in `servername`. A
   reader could reasonably worry that the IP substitution weakens TLS
   hostname verification. Document explicitly that Node's default
   `checkServerIdentity` matches the cert's SAN/CN against `servername`
   (not `host`), so a forged endpoint at the pinned IP with a valid
   cert for a different hostname would fail the handshake. No code
   change — the default behavior is correct; the comment forecloses
   future "this looks dangerous" review-round noise on the same line.

Items NOT addressed (with rationale documented elsewhere):

- Compliance Swarm V8.2.1 (retry route 404-vs-404 information leak):
  delivery IDs are UUIDs; the "leak" is the ability to probe existence
  of an opaque 128-bit identifier the caller already has, which is not
  meaningfully different from probing for any opaque token. Both
  branches return the same structured 404 envelope.

- Compliance Swarm CC7.2 (restore the .not() defense-in-depth filter):
  direct contradiction of last round's Greptile P2 fix. Greptile's
  PG-semantics analysis is correct — under READ COMMITTED, UPDATE
  re-evaluates WHERE against the row's current value when it acquires
  the lock, so .eq('status','in_flight') already handles the race.
  Adding a redundant .not() restores a misleading comment without
  closing a real gap. This is the documented Compliance Swarm
  oscillation pattern from the project's Phase 4 lessons.

- Compliance Swarm CC6.1 (webhook secret encryption-at-rest):
  architectural choice from PR-1; not in PR-3 (substrate hardening)
  scope. Belongs to a future hardening PR.

- Swedish-compliance review (operations queued/running rows hard-
  deletable): deliberate operability tradeoff — operators need to
  clear stuck/queued entries that crashed mid-flight. Blocking all
  deletes would force a manual DB intervention every time a worker
  crashed before reaching terminal status. The audit trail starts
  at terminal-state mutation, which IS blocked.

- Swedish-compliance review (salary_run.* / agi.* payload anonymisation
  after 7 years): already on the deferred-list as part of the 90-day
  TTL cleanup cron item from the PR description. Belongs to a
  retention-policy follow-up PR.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-15 20:38:19 +02:00