Files
accounted/lib/events/handlers/event-log-handler.ts
T
Jakob WennbergandClaude Fable 5 678f2ccffd feat(mcp): P2 hygiene — honest category suggestions, skill-reference lint, cadence copy (#882)
* feat(mcp): counterparty-tied category suggestions + no_signal (P2-1)

suggest_categories padded every transaction with a company-wide
category-frequency fallback at <=0.5 confidence — an identical four-way
spread on 20+/24 items that agents correctly reported as pure noise
(agent.feedback). Real signal came from memory atoms and query_journal.

- History is now counterparty-keyed: buildMerchantHistory groups past
  categorized transactions by normalized merchant; the engine only
  surfaces history for THIS transaction's merchant, with provenance
  ('Bokförd N gånger tidigare för denna motpart') and occurrence-scaled
  confidence (0.56 at 1x, capped 0.85). No global padding — an empty
  list is the honest answer.
- The MCP tool returns no_signal_transaction_ids for transactions where
  NO source matched, steering agents to investigate (query_journal)
  instead of pattern-matching on unrelated rows.
- Both callers (REST suggest-categories route + MCP tool) share the new
  helpers, so web UI and agents improve together.

Part of dev_docs/mcp_optimization_plan.md (P2-1).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* feat(skills): dangling-reference validation in skills:check + fix 10 dangling links (P2-2)

skills:generate/check now fail when an atom SKILL.md links a
references/*.md that does not exist on disk — a dangling pointer ships
a 404 to every agent that follows it (the weekly-booking-check
incident, agent.feedback).

The validator immediately caught 10 live dangling links in 4 atoms,
three distinct flavors:
- filename typo: swedish-asset-accounting/references/depreciaton.md
  renamed to depreciation.md (the link was right, the file misspelled)
- link mismatch: swedish-e-invoicing linked market-providers-pricing.md;
  the file is market-provider-pricing.md (link fixed)
- unauthored plans: single-shareholder-ab-fmb TODOs and reklambyra's
  'planerad utbyggnad' section used resolvable references/ paths for
  files that were never written — rephrased as plans without paths

Seed migration regenerated (4 atoms bumped, renamed reference child).

Part of dev_docs/mcp_optimization_plan.md (P2-2).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* docs(events): align agent-feedback review cadence copy (P2-4)

gnubok_feedback replies 'we aggregate signal weekly'; the event-log
handler comment said quarterly. One of them was lying — weekly wins
(the mcp_optimization_plan triage is the living example).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
2026-07-03 11:48:29 +02:00

210 lines
7.3 KiB
TypeScript

import { eventBus } from '@/lib/events/bus'
import type { CoreEventType } from '@/lib/events/types'
import { createServiceClientNoCookies } from '@/lib/auth/api-keys'
import { createLogger } from '@/lib/logger'
const log = createLogger('event-log')
/**
* Event types persisted to the event_log table for external automation platforms.
* Excludes noise events that are always followed by an actionable event.
*/
const PERSISTED_EVENT_TYPES: CoreEventType[] = [
'journal_entry.committed',
'journal_entry.reversed',
'journal_entry.corrected',
'document.uploaded',
'document.accessed',
'document.deleted',
'invoice.created',
'invoice.draft_deleted',
'invoice.sent',
'credit_note.created',
'transaction.synced',
'transaction.categorized',
'transaction.reconciled',
'period.locked',
'period.year_closed',
'customer.created',
'supplier.created',
'receipt.matched',
'receipt.confirmed',
'supplier_invoice.registered',
'supplier_invoice.approved',
'supplier_invoice.paid',
'supplier_invoice.credited',
'invoice.match_confirmed',
'supplier_invoice.match_confirmed',
'supplier_invoice.confirmed',
// MCP telemetry — every tool invocation, tools/list call, and resources/read.
// Lightweight metadata only. mcp.*/agent.* rows are retained 180 days by the
// cleanup cron (error-rate trends need more than the 30-day delivery window).
'mcp.tool_called',
'mcp.tools_list_called',
'mcp.resource_read',
// Workflow lifecycle + next-hint follow-through (Phase 3A). Tells us where
// agents stall, which skills actually drive completion, and whether the
// next-field rollout is paying off.
'mcp.workflow_started',
'mcp.workflow_completed',
'mcp.next_hint_followed',
// Every successful gnubok_load_skill, all tiers — which atoms agents
// actually load. Joined against mcp.tool_called error rates to measure
// whether a loaded atom helps or hurts.
'mcp.skill_loaded',
// Agent self-reported feedback — surfaces "this tool was missing", "this
// description was wrong", etc. Reviewed weekly (matches the gnubok_feedback
// reply copy); triage → dev_docs/mcp_optimization_plan.md.
'agent.feedback',
// Bank connection consent lifecycle — required audit trail per ASVS V16
// and GDPR Art.30 (records of processing) for PSD2 consent decisions.
'bank_connection.consent_granted',
'bank_connection.account_selection_changed',
'bank_connection.revoked',
'bank_connection.cash_account_mirror_failed',
]
// Excluded (with reasoning):
// - journal_entry.drafted: always followed by .committed
// - receipt.extracted: intermediate AI step; .matched/.confirmed are actionable
// - supplier_invoice.received: inbox receipt; .confirmed is actionable
// - supplier_invoice.extracted: intermediate AI step
/**
* Extract the primary entity ID from an event payload.
*/
function extractEntityId(payload: Record<string, unknown>): string | null {
// Try common entity shapes in priority order
const entityKeys = [
'entry', 'invoice', 'transaction', 'customer', 'supplier', 'receipt',
'supplierInvoice', 'creditNote', 'period', 'document', 'inboxItem',
] as const
for (const key of entityKeys) {
const entity = payload[key]
if (entity && typeof entity === 'object' && 'id' in entity) {
const id = (entity as Record<string, unknown>).id
if (typeof id === 'string') return id
}
}
// Flat-string ID fields on events that don't carry a full entity object.
// Bank connection events fall into this category — the connection lives in
// an extension table, so we record its id directly. invoice.draft_deleted
// carries only invoiceId because the row is already gone.
if (typeof payload.connectionId === 'string') {
return payload.connectionId
}
if (typeof payload.invoiceId === 'string') {
return payload.invoiceId
}
// For journal_entry.corrected: use the corrected entry's ID
if ('corrected' in payload) {
const corrected = payload.corrected
if (corrected && typeof corrected === 'object' && 'id' in corrected) {
const id = (corrected as Record<string, unknown>).id
if (typeof id === 'string') return id
}
}
// For journal_entry.reversed: use the reversal entry's ID
if ('reversalEntry' in payload) {
const reversalEntry = payload.reversalEntry
if (reversalEntry && typeof reversalEntry === 'object' && 'id' in reversalEntry) {
const id = (reversalEntry as Record<string, unknown>).id
if (typeof id === 'string') return id
}
}
return null
}
/**
* Strip userId and companyId from payload (each stored in its own column) and
* return clean data for the JSONB blob.
*/
function stripMetaFields(payload: Record<string, unknown>): Record<string, unknown> {
const { userId: _userId, companyId: _companyId, ...data } = payload
return data
}
/**
* Persist a single event to the event_log table.
*/
async function persistEvent(
eventType: string,
userId: string,
companyId: string,
entityId: string | null,
data: Record<string, unknown>
): Promise<void> {
const supabase = createServiceClientNoCookies()
const { error } = await supabase
.from('event_log')
.insert({
user_id: userId,
company_id: companyId,
event_type: eventType,
entity_id: entityId,
data,
})
if (error) {
log.error(`Failed to persist event ${eventType}:`, error.message)
}
}
/**
* Register event log handlers on the event bus.
* Persists events to the event_log table for external automation platforms.
* Returns an array of unsubscribe functions.
*/
export function registerEventLogHandler(): (() => void)[] {
return PERSISTED_EVENT_TYPES.map((eventType) =>
eventBus.on(eventType, async (payload) => {
try {
const rawPayload = payload as Record<string, unknown>
const userId = rawPayload.userId as string
const companyId = rawPayload.companyId
// All persisted event payloads mandate companyId in TypeScript; this guard
// protects against any caller that bypasses the type system. Writing NULL
// would make the row invisible to the company_id-scoped SELECT RLS policy.
if (typeof companyId !== 'string' || companyId.length === 0) {
log.error(`Event ${eventType} missing companyId; skipping persistence`)
return
}
// transaction.synced carries an array — batch insert
if (eventType === 'transaction.synced' && Array.isArray(rawPayload.transactions)) {
const transactions = rawPayload.transactions as Array<Record<string, unknown>>
if (transactions.length === 0) return
const rows = transactions.map(tx => ({
user_id: userId,
company_id: companyId,
event_type: eventType,
entity_id: typeof tx.id === 'string' ? tx.id : null,
data: { transaction: tx },
}))
const supabase = createServiceClientNoCookies()
const { error } = await supabase.from('event_log').insert(rows)
if (error) {
log.error(`Failed to persist batch transaction.synced:`, error.message)
}
return
}
const entityId = extractEntityId(rawPayload)
const data = stripMetaFields(rawPayload)
await persistEvent(eventType, userId, companyId, entityId, data)
} catch (err) {
log.error(`Event log handler error for ${eventType}:`, err)
}
})
)
}