Files

T

Mattsson 9ed0b9515a Fix/invoice booking vat fixes (#778 )

* feat(invoices): add Plusgiro input to bank details settings

Plusgiro was already persisted, validated by the API schema, rendered on
the invoice PDF and toggleable via "Visa plusgiro" — but the settings UI
had no field to enter the number, so plusgiro-only users could not fill
it in. Add the input next to Bankgiro with Luhn validation and hyphen
formatting, include it in the save payload (normalised on save so raw
digits still match the dashed schema format), and add sv/en strings.

Adds validatePlusgiroNumber/formatPlusgiroNumber helpers + tests.

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

* fix(invoices): respect non-VAT-registered seller in PDF preview + portal tooltips

Two user-reported bugs:

- PDF preview (/api/invoices/preview-pdf) ignored company.vat_registered and
  fell back to the customer-driven 25% rate, so a non-momsregistrerad seller
  saw VAT in the review step even though the created invoice books none. Mirror
  the server-side write gate (build-invoice-write.ts): force 0% when
  vat_registered is false (delivery notes excepted).

- InfoTooltip rendered TooltipContent without a Portal, so tooltips were
  clipped by the scrollable DialogContent (overflow-y-auto) in the send-invoice
  journal-entry review. Wrap in TooltipPrimitive.Portal.

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

* fix(transactions): book library mall from its literal lines, not a lossy fallback

Booking a bank transaction with a user-created booking-template (mall) via the
convertible "QuickReview" fast path reduced the template to a single category +
one account_override, silently discarding the chosen debit/credit. A
kundinbetalning mall (D 1930 / K 1510) booked as a generic cost (D 6991 / K 1930),
or with a VAT line as D 1930 / K 1930 / K 2611 — and the result flipped with the
direction inferred from the business/settlement line tags, so visually-identical
templates produced different verifikationer.

Route every library template through the journal-entry editor (applyTemplate ->
/book), which posts the literal lines, regardless of convertibility. Add
regression tests locking the contract.

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

* fix(bookkeeping): make the booking-time duplicate guard bypassable

TRANSACTION_BOOK_POSSIBLE_DUPLICATE told users they could "book anyway" but
the UI dead-ended on a toast with no way to do so. Add a shared
DuplicateBookingDialog that surfaces the already-booked sibling and lets the
user review it or book anyway (force bound to the reviewed candidate, which
the server re-detects so a stale id cannot wave the guard away).

- Wire the dialog into the /transactions categorize flow and the manual
  booking dialog (JournalEntryForm -> /api/transactions/[id]/book)
- Bind the override to expected_duplicate_transaction_id OR
  expected_duplicate_journal_entry_id so ledger-only vouchers (paid invoice,
  salary run) can be confirmed too
- Extend the guard to the pending-operations commit path and the MCP server
- Tests for book/categorize routes, detection, and the commit guard

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

* fix(bookkeeping): log duplicate-guard bypass to behandlingshistorik in the agent commit path

The web /book and /categorize routes append a durable
BankTransactionDuplicateDismissed event when a user books over a detected
possible double-booking. The agent commit path (commitCategorizeTransaction,
commitMarkInvoicePaid) skipped the guard silently on allow_duplicate=true,
leaving no behandlingshistorik — an auditor could not reconstruct why the
duplicate was allowed (BFNAR 2013:2 kap 8).

When allow_duplicate=true, re-detect the candidate and append the dismissal
event (BankTransactionDuplicateDismissed for the bank-line path,
InvoiceDuplicatePaymentDismissed for mark-paid). Best-effort — a logging
failure never blocks a legitimate booking. Payloads stay PII-safe (ids,
amounts, dates only — no customer or merchant name).

Also fix the misleading DuplicateBookingDialog JSDoc: the retry binds
expected_duplicate_journal_entry_id, not candidate.transaction_id, so the
systemdokumentation matches the actual control (BFL 7 kap).

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

* test(mcp-server): stub booking-duplicate guard in receipt-matcher categorize tests

The gnubok_categorize_transaction tool runs the booking-time duplicate guard
before staging; its detection queries consumed the queued supabase mock
results, so the staging assertions saw a thrown duplicate error instead of a
staged op. Mock detectBookingDuplicate to "no duplicate" since these tests
don't exercise that path.

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

* refactor(transactions): use roundOre for duplicate-guard öre rounding

Replace naive Math.round(x*100)/100 with roundOre() from @/lib/money in the
booking-time duplicate guard (detection lib, commit executor, MCP categorize
tool), satisfying the no-new-antipatterns ratchet guard.

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

---------

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

2026-06-25 15:54:35 +02:00

__tests__

Fix/invoice booking vat fixes (#778 )

2026-06-25 15:54:35 +02:00

prompts

Accounted rebrand + swarm-skill cleanup + bank-reconciliation fixes (#643 )

2026-06-03 10:52:01 +02:00

resources

Accounted rebrand + swarm-skill cleanup + bank-reconciliation fixes (#643 )

2026-06-03 10:52:01 +02:00

skills

feat(mcp): self-describing agent surface — staging _meta, company identity, clean skill summaries (#775 )

2026-06-25 13:42:12 +02:00

widgets

Accounted rebrand + swarm-skill cleanup + bank-reconciliation fixes (#643 )

2026-06-03 10:52:01 +02:00

index.ts

fix(mcp): return 405 (not 401) on GET /mcp to stop client re-auth storm (#747 )

2026-06-17 14:26:13 +02:00

manifest.json

feat: MCP server, API keys, OAuth, and KPI dashboard (#72 )

2026-03-21 12:57:14 +01:00

origin-guard.ts

feat(mcp): Origin-header validation + serverInfo title + connect-claude docs export (P0-4 follow-up) (#684 )

2026-06-08 09:14:46 +02:00

README.md

feat(mcp): tool titles + Connect-with-Claude docs for Connectors Directory readiness (P0-4) (#682 )

2026-06-06 11:47:21 +02:00

server.ts

Fix/invoice booking vat fixes (#778 )

2026-06-25 15:54:35 +02:00

tool-result.ts

Add/ai native supp (#385 )

2026-05-04 11:12:29 +02:00

README.md

Accounted MCP server

JSON-RPC 2.0 server exposing the Accounted bookkeeping engine to MCP clients (Claude Desktop, Claude Code, etc.). Endpoint: /api/extensions/ext/mcp-server/mcp. OAuth and stdio bridge live alongside the API surface — see app/api/mcp-oauth/ and packages/gnubok-mcp/.

Tool authoring contract

Enforced by tests in __tests__/ — these are not style preferences, they're guard rails.

additionalProperties: false on every inputSchema. Guarded by strict-schemas.test.ts. Forces clear rejections on hallucinated fields instead of silent ignores.
Descriptions ≤ 280 chars. Guarded by output-schema.test.ts. No Args: / Returns: / Examples: prose — those belong in JSON Schema. Use agent-native hints ("Use to…", "Call X first", "HIGH risk").
Staged-operation envelope for write tools — outputSchema: STAGED_OPERATION_SCHEMA (server.ts). Fields: staged, risk_level, actor, message, preview, period_status?, next?. The staged: true boolean is the explicit completion signal; agents must not infer completion from prose. Do NOT introduce a parallel { success, shouldContinue, output } envelope.
period_status threading — any tool that ties to a fiscal-period-bound date (categorize, mark paid, create voucher, correct/reverse entry, approve supplier invoice) passes dateForPeriodCheck to stagePendingOperation. Response then includes period_status: { period_id, status: open|locked|closed, lock_date } so widgets and agents disable writes without round-trips.
Scope mapping — every new tool needs an entry in lib/auth/api-keys.ts TOOL_SCOPE_MAP. Missing entries default to deny.
Tests for new write tools — add staging-gate coverage to __tests__/voucher-tools.test.ts (or a sibling) plus executor coverage to lib/pending-operations/__tests__/voucher-executors.test.ts if the tool stages a new operation_type.

Determinism / cache stability

Tool definitions (name, description, inputSchema, outputSchema, annotations) are declared as static object literals at module load — no timestamps, no UUIDs, no Date/Math.random in the definition layer. This makes the tools/list JSON payload byte-stable across requests, which lets agent-side prompt caches stay warm. Do not introduce per-request non-determinism into the definitions block. Anything time-bound or random belongs inside execute().

For internal Anthropic API usage (today only extensions/general/invoice-inbox/lib/extract-invoice-fields.ts): annotate stable prefixes with cache_control: { type: 'ephemeral' } and log usage.cache_read_input_tokens for hit-ratio observability. The 1h TTL from the agent-native API plan (item 10) requires the direct Anthropic API; Accounted's Bedrock path defaults to a shorter TTL.

Payload-size watchdog

payload-size.bench.test.ts enforces a tools/list JSON payload ceiling (currently 36,000 tokens — bumped from 32,000 when top-level Tool.title landed on all tools for Claude Connectors Directory readiness). If the test fires, the right answer is rarely "raise the ceiling" — instead, trim descriptions or leverage gnubok_search_tools (already deployed; tool definitions can defer to it for discovery rather than enumerating in tools/list).

Where things live

server.ts — the tools array + JSON-RPC dispatcher
tool-result.ts — withNext(), toToolError() response helpers
resources/ — read-only Accounted:// URIs (active company, period, recent activity, capabilities, attention items, voucher gaps, chart of accounts, VAT treatments)
widgets/ — inline HTML widgets (receipt-matcher, vat-review)
prompts/ — slash-command-style prompts
skills/ — domain-knowledge skill bodies served via gnubok_load_skill
__tests__/ — strictness guards + per-tool coverage