ccdfed5fea
* 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>
gnubok MCP server
JSON-RPC 2.0 server exposing the gnubok 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: falseon everyinputSchema. Guarded bystrict-schemas.test.ts. Forces clear rejections on hallucinated fields instead of silent ignores.- Descriptions ≤ 280 chars. Guarded by
output-schema.test.ts. NoArgs:/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?. Thestaged: trueboolean is the explicit completion signal; agents must not infer completion from prose. Do NOT introduce a parallel{ success, shouldContinue, output }envelope. period_statusthreading — any tool that ties to a fiscal-period-bound date (categorize, mark paid, create voucher, correct/reverse entry, approve supplier invoice) passesdateForPeriodChecktostagePendingOperation. Response then includesperiod_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.tsTOOL_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 tolib/pending-operations/__tests__/voucher-executors.test.tsif the tool stages a newoperation_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; gnubok's Bedrock path defaults to a shorter TTL.
Payload-size watchdog
payload-size.bench.test.ts enforces a tools/list JSON payload ceiling (currently 25,000 tokens). 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 dispatchertool-result.ts—withNext(),toToolError()response helpersresources/— read-onlygnubok://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 promptsskills/— domain-knowledge skill bodies served viagnubok_load_skill__tests__/— strictness guards + per-tool coverage