c84f951a5c2f51279277a506754b2bd94de28249
15
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
3b3adf96c7 |
feat(invoice-inbox): receipt-aware extraction + OCR pipeline fixes (#1331)
Receipts and invoices were extracted through one invoice-shaped prompt with no document classification. The extractor now also returns documentKind, payment method (+ card last4), purchaseTime, merchantCategory and legibility, validated with .catch(null) so a hallucinated label degrades to unknown instead of sinking the parse. The FieldsRail shows type and payment method above the editable fields. Pipeline fixes, all verified against real failure paths: - PDFs >3 pages: extract from a pdf-lib slice of the first 3 pages instead of skipping entirely (issue #553 gate); truncation recorded in extracted_data.pages and shown in the UI. - Oversized images (>4 MB, over Bedrock's 5 MB cap): downscale to <=2000px JPEG via sharp before base64, instead of erroring to an empty result. - HEIC/HEIF: attempt sharp transcode to JPEG; when libvips lacks HEIF (prebuilt binaries), fall through to today's behavior but show an explicit hint instead of silently blank fields. - Oresavrundning: prompt rule + totals.roundingAmount so receipt totals reconcile with subtotal+VAT for exact-amount transaction matching. - retry-extraction overwrites extracted_data wholesale: a confirm now guards against silently destroying manual field edits. Deliberately NOT added: retry-on-transient-Bedrock-error; the SDK already retries twice by default (maxRetries=2). Co-authored-by: Jakob Wennberg <jakob.wennberg@gmail.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
Jakob WennbergandClaude Opus 5
|
2d543ac999 |
feat(agent): move every model call to Sonnet 5 (#1218)
* feat(agent): move every model call to Sonnet 5
Sonnet 5 is verified enabled on our Bedrock account already: a live probe of
eu.anthropic.claude-sonnet-5 in eu-north-1 answered normally, so no model-access
request was needed. The bare anthropic.claude-sonnet-5 is rejected (on-demand
throughput needs the cross-region inference profile), so the eu. prefix we
already use stays.
This is not a model-string swap. Sonnet 5 REJECTS the fixed thinking budget
outright: thinking {type:'enabled', budget_tokens} returns 400 "not supported
for this model. Use thinking.type.adaptive and output_config.effort". Every
chat intent set a budget, so the assistant would have failed on the first turn
after a bare ID change. Reasoning depth is now an effort level (STANDARD high,
DEEP xhigh), and max_tokens is explicit per tier rather than derived from a
budget that no longer exists.
display:'summarized' is load-bearing, not cosmetic. The default is 'omitted',
which still emits thinking blocks but with empty text. Measured on our own
account at xhigh effort: summarized returned ~1k characters of reasoning, the
default returned none. Without it the collapsible "Tänker ..." block in the
chat would have gone silently empty, which no mocked test would have caught.
Ceilings are raised (16k standard, 24k deep) because Sonnet 5's tokenizer
produces roughly 30% more tokens for the same text and max_tokens now caps
thinking and the visible reply together.
Also resolves the Opus 4.7 landmine recorded in the readiness doc: the composer
comment told ops to flip BEDROCK_OPUS_MODEL_ID to Opus 4.7, which would have
400d every thinking intent against the legacy budget shape. Both model
constants now point at Sonnet 5 and the stale instruction is gone.
Checked but deliberately unchanged: forced tool_choice in atom-selection. The
Sonnet 5 docs require thinking:{type:'disabled'} alongside a forced tool_choice
on Bedrock; probed against our account, the forced call succeeds without it, so
no change was made rather than adding a guard we cannot show is needed.
Other call sites moved too: invoice-inbox extraction, document extraction, the
compliance config, and the CI/CD workflows (pr-agent MODEL and MODEL_WEAK,
swedish-compliance-review, compliance-swarm).
Verified: 11315 tests pass, lint and tsc clean on every touched file, guards
pass.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
* fix(agent): review triage: keep the no-thinking output ceiling, finish the model sweep
max_tokens now caps thinking and the visible reply together, so collapsing the
two tiers into one made every non-thinking intent inherit a 16000 ceiling where
it used to have 4096. Give it its own MAX_TOKENS_NO_THINKING instead, set to the
old 4096 scaled ~30% for Sonnet 5's tokenizer so the effective reply length is
unchanged rather than quietly cut.
scripts/swedish-compliance-review.mjs still fell back to Sonnet 4.6 when
REVIEW_MODEL was unset, so a manual run silently used the old model. The initial
sweep only covered .ts and .yml.
pr-agent's FALLBACK_MODELS listed the primary model as its own fallback, which is
not a fallback; dropped it and rewrote the surrounding comments, which still
described Opus 4.8 and a 200k window.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
||
|
MattssonandClaude Fable 5
|
3a3c4adbc6 |
Bug/ai assistant config (#939)
* revert(agent): restore plain AWS_* Bedrock credential handling
Undoes the credential-name change from #937 (
|
||
|
Mattsson
|
ae489cfdcb | fix(client): update AWS credential handling to prefer BEDROCK_AWS_* environment variables (#937) | ||
|
Jakob WennbergandClaude Sonnet 5
|
ec27228a8e |
style: remove em/en dashes repo-wide, add CLAUDE.md rule against them (#890)
Em dashes (—) and en dashes (–) had spread across comments, docs, tests, and a few UI strings, reading as AI-generated boilerplate rather than house style. Replaced each with punctuation matching its context: colon for explanatory clauses, comma for asides, plain hyphen for numeric/legal ranges (e.g. "21-23§"), "to"/"till" for date ranges, parentheses for paired-dash asides. messages/en.json and messages/sv.json were fixed by hand together to keep sv/en in sync. Left untouched where the dash is the functional subject rather than decorative punctuation: date-range-parser.ts's separator regex, charset-repair.ts's CP1252 byte-mapping table (and its test), the SIE encoding mojibake docs, generic-csv.ts's minus-sign normalizer, the agent system-prompt files that already instruct against em dashes, and a golden iXBRL test fixture compared byte-for-byte. Also fixes two bugs surfaced along the way: an off-by-one in ApiKeysPanel's scope-label split (a leftover from an earlier partial pass), and a charset-repair test that had lost the literal en-dash it exists to verify. Regenerated the agent atom seed migration (skills:generate) since 27 SKILL.md files changed. Added a CLAUDE.md rule against em/en dashes, with an explicit carve-out for the functional-dash cases above. Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com> |
||
|
Jonas FlodénandClaude Sonnet 4.6
|
d5abc8dd1c |
fix(inbox): set_inbox_extracted_data should accept and persist accountSuggestion (#760)
The tool was using InvoiceExtractionSchema which forces every
lineItems.accountSuggestion to null via .transform() — the same guard
that prevents the AI extractor from hallucinating BAS accounts.
Agents supplying their own extraction should be able to pin a cost
account per line.
Adds AgentExtractionSchema (exported alongside ExtractionSchema)
where accountSuggestion accepts a validated BAS expense account
(class 4–7, /^[4-7]\d{3}$/) or null. set_inbox_extracted_data now
parses through this schema so the field survives the round-trip to
the DB and is available when gnubok_create_supplier_invoice_from_inbox
builds line items.
Signed-off-by: Jonas Flodén <jonas@floden.nu>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
|
||
|
MattssonandClaude Fable 5
|
db8983ba9e |
Add/bokslut (#718)
* feat(arcim-migration): Briox provider with SIE-over-API import - Briox auth via account ID + application token (no app-level credentials); both tokens rotate on refresh and are persisted - New sie-fetcher pulls the general ledger as SIE through the provider API for Fortnox, Briox and Bjorn Lunden - Wizard stops on a failed SIE import and surfaces the real errors instead of proceeding to the misleading migrate-guard message - PROVIDER_SIE_ONLY_FORTNOX renamed to PROVIDER_SIE_NOT_SUPPORTED; new PROVIDER_TOKEN_INVALID for rejected provider credentials Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(bookkeeping): per-line accruals (periodisering) on invoices and supplier invoices Defer revenue/costs per invoice line to 29xx/17xx interim accounts with automatic monthly dissolution (nightly cron + catch-up at registration), schedule cancellation on credit, year-end auto-detect exclusion for already-scheduled invoices, invoice-inbox service-period extraction for prefill, and an MCP tool to list schedules. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * feat(bokslut): iXBRL arsredovisning generation and Bolagsverket digital filing Generate the annual report as iXBRL from a generated taxonomy registry (K2 element lists, taxonomy:generate/check scripts + CI guard), expose it via the fiscal-period API, and add the bolagsverket extension for digital submission to eget utrymme with webhook-driven status tracking (submissions table + pg tests, lifecycle events, year-end wizard UI). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * test(mcp): raise origin-guard test timeout to 20s The dynamic import pulls in the full server module; the parse alone flirts with the 5s default under full-suite parallel load. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Add new scripts and documentation for K2 AB taxonomy generation and validation - Introduced `generate-taxonomy-registry.ts` to automate the generation of the iXBRL taxonomy concept registry from official element lists and tuple models. - Added `validate-ixbrl.mjs` for validating generated iXBRL reports against the official taxonomy package using Arelle. - Included new documentation files: - `k2-ab-arsredovisning-elementlista-2024-09-12_rev20250312_sv.xlsx` - `tuple-innehallsmodell-arsredovisning-k2-2024-09-12.xlsx` - `taxonomi-paket-2024-09-12_rev20250312.zip` * Add tests for bookkeeping accruals dissolution and supplier invoices - Implement tests for the POST /api/bookkeeping/accruals/[id]/dissolve route, covering success and error scenarios. - Add tests for the DELETE /api/supplier-invoices/[id] route, including authentication checks and validation of invoice deletion conditions. - Introduce tests for the Arcim migration provider client, ensuring token handling and error classification. - Create tests for the Bolagsverket extension, validating submission role enforcement and environment settings. - Add Zod schemas for Bolagsverket response payloads to ensure proper validation. - Implement tests for MCP server's list accrual schedules, confirming registration and scope mapping. - Add consistency tests for IXBRL document generation, ensuring duplicate facts and XML escaping are handled correctly. - Introduce typed domain errors for accrual schedules to improve error handling in the service. - Add tests for resolving consent with Briox token refresh concurrency, ensuring proper token management and error handling. * fix(tests): update payload size guard comments to reflect recent changes in tool descriptions and ceiling adjustments * fix(gitattributes): mark generated JSON files in bokslut taxonomy as linguist-generated * feat(migrations): add backfill for invoices.journal_entry_id and fallback for next_voucher_number user_id * feat(bokslut): enhance compliance and financial processing features with new submission details and security measures --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|
Jakob WennbergandClaude Opus 4.8
|
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>
|
||
|
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. |
||
|
Jakob WennbergandClaude Opus 4.7
|
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>
|
||
|
Jakob WennbergandClaude Opus 4.7
|
8a1c7f0714 |
feat(invoice-inbox): loading state + review fixes + env-driven config (#416)
* feat(invoice-inbox): make Bedrock model + max_tokens env-overridable Read BEDROCK_MODEL_ID and BEDROCK_MAX_TOKENS from process.env so ops can swap models or raise token caps without a code deploy. Defaults align with production: eu.anthropic.claude-sonnet-4-6 in eu-north-1, 8192 tokens (up from 4096 — enough headroom for 20+ line-item invoices that were still occasionally truncating mid-JSON). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(invoice-inbox): guard onFieldsUpdated, tighten env parsing, hoist afterAll Three fixes from Greptile review on #416: 1. Stale-closure bug in onFieldsUpdated. setSelected used an unconditional spread, so a PATCH that completed after the user had navigated to a different item would silently overwrite the new selection's extracted_data with the old item's payload. Guard with prev?.id === selected.id (capturing selected.id outside the updater so it's stable through both setSelected and setItems). 2. BEDROCK_MAX_TOKENS env parser used `Number(...) || 8192`, which silently treats a deliberate `0` the same as unset. Use Number.isFinite(n) && n > 0 so an invalid configuration is not masked by the fallback. 3. Hoist `afterAll` import in the test file to the top alongside other vitest imports. Was working via ES-module hoisting but reads weird to humans and linters. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|
Jakob WennbergandClaude Opus 4.7
|
9c13586b13 |
feat(invoice-inbox): AI extraction (Bedrock Sonnet 4.6) + editable fields (#415)
* feat(invoice-inbox): AI extraction via Bedrock + editable fields Two big changes that go together: 1. Replace the regex extractor with Claude Sonnet 4.6 via AWS Bedrock. The PDF or image is sent directly to the model — no unpdf, no DOM stubs, no worker bundling. Sonnet handles English receipts (Anthropic, AWS, Stripe), USD/EUR currency symbols, scanned PDFs and image receipts, and Subtotal vs Total disambiguation that the regex layer couldn't. Output is JSON, validated with Zod; anything that doesn't parse falls back to an empty result so the inbox row still lands. 2. Make the extracted fields editable inline in the workspace. Each field is an Input bound to local draft state with debounced auto-save (800ms) to a new PATCH /items/:id/fields route. The route refuses edits once the item is converted to a supplier invoice. The parent updates both the selected-item view and the list rail when a field saves so the summary stays in sync. Drops `unpdf`, adds `@anthropic-ai/bedrock-sdk`. Requires three env vars in Vercel prod (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION) — not yet set; when missing, the extractor logs a warning and returns empty so the upload still works. All 43 invoice-inbox unit tests pass. tsc clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * feat(invoice-inbox): loading state + #415 review fixes Loading state for AI extraction: - Optimistic placeholder row inserted into the workspace list the moment an upload starts. Shows the file name + a "Tolkar dokument med AI…" spinner instead of timestamp/amount. - Document preview pane shows a centered loader + same caption while the placeholder is selected. - Fields rail shows 6 skeleton inputs + the same caption. - Action buttons are hidden until the real row arrives. - Placeholder is removed on success or failure (toast on error). PR #415 review follow-ups (Greptile + Swedish compliance bot): - Bump max_tokens 1500 → 4096 so multi-line invoices don't silently truncate mid-JSON and land with all-null fields. - Tighten NullableDate to range-checked regex + Date.parse refine — rejects 2026-13-45 / 2026-02-30 with a clean 400 instead of a 500. - Make vatRate representation consistent: percent integer (25, 12, 6, 0) for both lineItems[].vatRate AND vatBreakdown[].rate. Previously the AI was instructed to emit decimals for one and integers for the other. - EditableFieldsList re-seeds drafts when the parent passes a new data snapshot, but only on fields where the local draft still matches the previous server value — so a server-normalised currency upper-cases cleanly without clobbering an in-progress edit. - 409 conflict (item already linked to a supplier invoice) shows the server's specific Swedish message ("Posten är låst") instead of the generic "Kunde inte spara". Out of scope (separate tickets if needed): live updates for email-arrived items, AWS_SESSION_TOKEN plumbing, 6 % livsmedel rate transition validation, server-side totals consistency check. All 43 invoice-inbox unit tests pass. tsc clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(invoice-inbox): tighten extraction schema per compliance bot review Three follow-ups from the Swedish compliance bot's second pass on #415: 1. lineItems[].vatRate and vatBreakdown[].rate now refine to 0–100, blocking AI hallucinations like 5000% or negative rates while still accepting non-Swedish rates (UK 20, DE 19) since gnubok stores foreign invoices for reference. The strict Swedish [0, 6, 12, 25] allowlist is not enforced here on purpose — that check belongs in the supplier- invoice-creation step where the data hits the ledger. 2. accountSuggestion is now coerced to null at parse time via .transform, eliminating the brief intermediate window where a hallucinated string could appear in the parsed object before the post-validation .map() nulled it. Removes the redundant .map() afterwards. 3. PATCH /items/:id/fields currency now requires ISO 4217 format (^[A-Z]{3}$). Previously accepted any 3–8 char string, which would flow into supplier-invoice creation and produce a faktura with an invalid currency. Test fixture vatRate updated 0.25 → 25 to match the percent-integer convention introduced in the previous commit. All 43 invoice-inbox 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> |
||
|
Jakob WennbergandClaude Opus 4.7
|
8676f10fdd |
fix(invoice-inbox): switch from pdfjs-dist to unpdf (#409)
* fix(invoice-inbox): switch from pdfjs-dist to unpdf for PDF text extraction After three rounds of fighting pdfjs on Vercel (#407 stubbed DOM globals, #408 tried to ship the worker file via outputFileTracingIncludes), text extraction still failed in prod with "Setting up fake worker failed" — Next's tracer can't reliably include pdfjs-dist's worker file when the package is marked as a server external. unpdf is a serverless-first wrapper around pdfjs (by unjs) that ships its own bundled pdfjs build with no canvas/worker dependencies. Drop-in replacement: extractText returns merged page text directly. - Remove DOM stubs, serverExternalPackages, outputFileTracingIncludes - Replace pdfjs-dist with unpdf (no transitive deps) - Update test mock from getDocument → extractText All 45 invoice-inbox unit tests pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(invoice-inbox): make unpdf import static, rename pdfjs test labels Per Greptile review on #409: 1. CRITICAL: tryExtractPdfText was still using await import('unpdf'), a dynamic import. CLAUDE.md forbids dynamic imports in extensions precisely because Next.js bundling can't reliably trace them — which is the same class of failure that caused the pdfjs prod bug. unpdf bundles statically (no canvas/worker), so a top-level static import is safe and correct. 2. NIT: two test descriptions still said "pdfjs" after the mock rename. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|
Jakob WennbergandClaude Opus 4.7
|
5f9f72959f |
fix(invoice-inbox): stub DOM globals so pdfjs runs on Vercel (#407)
* fix(invoice-inbox): stub DOM globals so pdfjs text extraction works on Vercel pdfjs-dist's legacy build references DOMMatrix/ImageData/Path2D at module load. On Vercel's Node runtime those globals don't exist, so the dynamic import threw "DOMMatrix is not defined" before getDocument() ran — every inbox item came back with empty extracted_data. We only call getTextContent (no rendering), so empty-class stubs are sufficient. Also mark pdfjs-dist as a server external package so Next.js doesn't try to bundle it (the bundling step pulls in @napi-rs/canvas references that aren't installed and fails silently with the same downstream error). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * refactor(invoice-inbox): hoist DOM stubs to module scope Per Greptile review on #407: stubs run once at module load instead of per-call. Same effect, clearer intent. Wrapped in a block to scope the local variable. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> |
||
|
Mattsson
|
fa7d4075cf |
Supp/invoice bfl errors (#390)
* feat(accounting): update accounting method validation and messaging for aktiebolag and enskild firma * Remove AI subsystem and related code - Deleted AI proposals and requests persistence logic from `lib/ai/proposals/persist.ts`. - Removed re-validation logic for proposals in `lib/ai/proposals/re-validate.ts`. - Cleaned up schemas related to AI flows in `lib/api/schemas.ts`. - Removed AI-related fields from bookkeeping engine in `lib/bookkeeping/engine.ts`. - Eliminated AI event types from `lib/events/types.ts`. - Updated tests to reflect the removal of AI-related functionality in `lib/extensions/__tests__/sectors.test.ts`. - Adjusted initialization logic in `lib/init.ts` to exclude AI proposal handler registration. - Cleaned up transaction ingestion logic in `lib/transactions/ingest.ts` to remove AI flow checks. - Updated helper functions in `tests/helpers.ts` to remove AI-related settings. - Removed AI-related types and interfaces from `types/index.ts`. - Added migration script to drop AI-related tables and settings from the database. * fix(migrations): ensure foreign key constraint is dropped before removing AI tables * feat(invoice-inbox): implement deterministic invoice field extraction and inbox provisioning - Added `extract-invoice-fields.ts` for extracting fields from PDF invoices using regex and pdfjs-dist, replacing the previous AI classifier. - Introduced `inbox-provisioning.ts` to manage company inbox addresses and rotation of inboxes using Supabase RPCs. - Created `resend-inbound.ts` for handling inbound email events and attachments via the Resend API. - Defined the extension manifest for the invoice inbox, specifying required environment variables and descriptions. - Migrated database schema to remove AI-related columns and tighten the status enum in `invoice_inbox_items`. * feat(invoice-inbox): remove AI-specific columns and tighten status enum * fix(skattekonto): remove manual entry creation reference from transaction input * fix(schemas): remove accounting method validation for aktiebolag in UpdateSettingsSchema |