Open-source manufacturing ERP, MES and QMS. Quoting, MRP, inventory, shop floor, quality and lot/serial traceability on one Postgres schema, with a REST API and MCP server. Self-host or use Carbon Cloud.
TypeScript ai-agentsbill-of-materialserpinventory-managementleanmanufacturingmanufacturing-execution-systemmcp-servermesmrpopen-source-erppostgresqlproduction-planningqmsquality-management-systemreact-routerself-hostedsupabasetraceabilitytypescript
// SPDX-License-Identifier: AGPL-3.0-only
// Carbon (github.com/crbnos/carbon). Modified or adapted versions of this file,
// including ports, remain AGPLv3; serving them over a network requires releasing their source.
/**
* Every event id a live workflow subscribes to must still resolve. Needs a live database:
* `pnpm --filter @carbon/checks workflow-events`. Does NOT run in CI — no job in
* .github/workflows has database credentials. Run it by hand after a catalog change.
*
* Two shapes of drift:
* - a shipped id the catalog no longer has (a rename left it behind);
* - a custom-field id whose `customField` row is gone. That one the catalog cannot
* catch on its own — it PARSES those ids rather than looking them up, deliberately,
* so that it stays company-blind. The existence check therefore lives here.
*/
import { createWorkflowCatalog, REGISTRY_ENTRIES } from "@carbon/ee/workflows";
import { Client } from "pg";
const CUSTOM_FIELD_EVENT =
/^([A-Za-z][A-Za-z0-9]*)\.customFields\.([^.]+)\.changed$/;
async function main() {
const connectionString =
process.env.DATABASE_URL ?? process.env.SUPABASE_DB_URL;
if (!connectionString) {
console.error(
"Set DATABASE_URL (or SUPABASE_DB_URL) to the Postgres connection string."
);
process.exit(2);
}
const catalog = createWorkflowCatalog();
const client = new Client({ connectionString });
await client.connect();
try {
const { rows } = await client.query<{
eventId: string;
workflowId: string;
companyId: string;
}>(
`SELECT DISTINCT t."eventId", t."workflowId", t."companyId"
FROM "workflowTriggerEvent" t
JOIN "workflow" w
ON w."id" = t."workflowId" AND w."companyId" = t."companyId"
WHERE w."publishedVersionId" IS NOT NULL
ORDER BY t."eventId"`
);
// `${companyId}:${table}:${fieldId}` for every custom field that still exists.
const { rows: fields } = await client.query<{
id: string;
table: string;
companyId: string;
}>(`SELECT "id", "table", "companyId" FROM "customField"`);
const known = new Set(
fields.map((f) => `${f.companyId}:${f.table}:${f.id}`)
);
const failures: string[] = [];
for (const row of rows) {
if (catalog.getEvent(row.eventId) === undefined) {
failures.push(
`FAIL workflow ${row.workflowId} (company ${row.companyId}) subscribes to "${row.eventId}", which is not in the catalog.`
);
continue;
}
// The catalog only proves the id is well formed and names a triggerable entity.
// Whether the field itself still exists is company data, so it is checked here —
// this is what surfaces a trigger left behind by a deleted custom field.
const match = CUSTOM_FIELD_EVENT.exec(row.eventId);
if (match === null) continue;
const [, entity, fieldId] = match;
const table =
entity === undefined ? undefined : REGISTRY_ENTRIES[entity]?.table;
if (table === undefined || fieldId === undefined) continue;
if (!known.has(`${row.companyId}:${table}:${fieldId}`)) {
failures.push(
`FAIL workflow ${row.workflowId} (company ${row.companyId}) subscribes to custom field "${fieldId}" on ${table}, which no longer exists in this company.`
);
}
}
for (const line of failures) console.log(line);
if (failures.length === 0) {
console.log(
`PASS workflow-events (${rows.length} subscription(s) all resolve)`
);
}
process.exit(failures.length > 0 ? 1 : 0);
} finally {
await client.end();
}
}
void main();