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.
/**
* Pure inspection verdicts shared by the engine (`@carbon/database/quality`) and
* the dataset seed, so a seeded sample carries exactly the status the engine
* would derive. No I/O and no kysely, so it loads in any runtime.
*/
export type InspectionVerdict = "Pending" | "Passed" | "Failed";
export type MeasurementFeatureSpec = {
type: string;
nominalValue: string | null;
tolerancePlus: string | null;
toleranceMinus: string | null;
};
export function parseSpecNumber(
value: string | null | undefined
): number | null {
if (value == null) return null;
const trimmed = value.trim();
if (trimmed === "") return null;
const parsed = Number(trimmed.replace(/^\+/, ""));
return Number.isNaN(parsed) ? null : parsed;
}
// Measurement features with a parseable nominal are judged numerically inside
// [nominal - |tol-|, nominal + |tol+|]; everything else (attribute features,
// GD&T strings that don't parse) is a pass/fail toggle.
export function valuateMeasurement(
feature: MeasurementFeatureSpec,
value: number | null,
passed?: boolean | null
): InspectionVerdict {
const nominal =
feature.type === "Measurement"
? parseSpecNumber(feature.nominalValue)
: null;
if (feature.type === "Measurement" && nominal !== null) {
if (value == null) return "Pending";
const tolPlus = Math.abs(parseSpecNumber(feature.tolerancePlus) ?? 0);
const tolMinus = Math.abs(parseSpecNumber(feature.toleranceMinus) ?? 0);
return value >= nominal - tolMinus && value <= nominal + tolPlus
? "Passed"
: "Failed";
}
if (passed == null) return "Pending";
return passed ? "Passed" : "Failed";
}
// Sampling is count-based, not positional — a feature's n is the minimum
// number of readings across ANY samples (per-feature gating at disposition
// enforces the counts), so a sample's own verdict is: Failed the moment any of
// its readings fails, Passed once every lot feature has a passing reading on
// it (a fully-inspected unit), otherwise Pending. A lot with no features never
// passes a sample.
export function deriveSampleStatus(
lotFeatureIds: readonly string[],
measurements: readonly { inspectionFeatureId: string; status: string }[]
): InspectionVerdict {
if (measurements.some((m) => m.status === "Failed")) return "Failed";
const allFeaturesPassed =
lotFeatureIds.length > 0 &&
lotFeatureIds.every(
(id) =>
measurements.find((m) => m.inspectionFeatureId === id)?.status ===
"Passed"
);
return allFeaturesPassed ? "Passed" : "Pending";
}
// Terminal states (Passed/Failed/Partial) are owned by the disposition path,
// so the per-sample recompute only flips between Pending and In Progress.
export function computeLotStatus(
samples: readonly { status: string }[]
): "Pending" | "In Progress" {
return samples.some((s) => s.status !== "Pending")
? "In Progress"
: "Pending";
}