A QIF-native quality and metrology platform, delivered as a Grafana app. It manages the whole inspection lifecycle - parts, GD&T characteristics, plans, devices, results and non-conformances - on a domain model that is QIF 3.0 all the way down.
Not generally available yet - this page describes what is built
QIF geometry feature types
GD&T characteristic types, ASME and ISO
reference files, round-trip tested
FAI, PPAP and APQP metadata built in
Most inspection software stores data in its own schema and offers QIF as an export filter. The export can only carry what that schema holds. Anything it has no field for goes missing on the way out, and the person receiving the file has no way of knowing what.
Here the domain model is QIF 3.0. Parts and assemblies, 33 geometry feature types - circles, cylinders, planes, threads, welds, point clouds - and 52 GD&T characteristic types across ASME and ISO. Import and export both move through XSD-validated QIF 3.0 XML, round-trip tested against the NIST reference files.
In practice that means a file survives the trip out and back, and a supplier's QIF can be imported as it arrives.
The inspection lifecycle end to end, from the drawing to the non-conformance and back.
Drag in a QIF or XML file and get a parsed summary - part, features, characteristics, saved views, plan, results. Preview and select row by row, then approve to import, with safe re-import and full ID reconciliation. Anything the parser does not natively model round-trips losslessly through a pass-through on export.
Ordered, reorderable steps, each tied to specific characteristics, a measurement device and a sampling strategy - every part, AQL, fixed interval or skip-lot. Plans move through draft, submit, approve or reject, and deprecate, with attachments on any step.
Two guided paths against a plan: a full desktop flow, and a kiosk mode built for the floor - large touch targets, a numeric keypad, QR and barcode scanning for serial or lot ID, and automatic pass/fail evaluation as the numbers go in.
Every part gets an interactive 3D viewport with an assembly tree, GD&T annotations laid over the model and saved camera views. A companion geometry service converts embedded QIF B-rep and STEP/IGES into web-viewable geometry, so the inspector works against the actual part.
Measurement sessions carry the metadata regulated manufacturers need as part of the session record itself - inspector, device, order and report numbers, and FAI, PPAP or APQP mode.
A registry for CMMs, articulated-arm CMMs, laser trackers, CT scanners, microscopes and calipers, with calibration tracking throughout - valid, due soon or overdue, plus history, certificates and calibration-master references.
A full NCR lifecycle - open, under review, dispositioned, closed, reopened - with severity, disposition (use as is, rework, scrap, return to vendor, waiver, re-inspect), root-cause categorisation, CAPA tracking, links back to the measurements that triggered it, a status history audit trail and threaded discussion. Raise one in a click from a failed result.
Because it is a Grafana app, the home screen is a real cockpit on Grafana's own engine - rolling pass rate, today's inspection count, open NCRs, a 14-day trend, per-device panels and overdue-measurement alerts, live against the operational database and configurable per organisation.
Multi-tenant by design, one tenant per Grafana organisation, provisioned automatically on first use - with object storage for every attachment and CAD asset, an immutable audit log, and per-tenant licensing built in from the start.
The usual response to AI in quality software is a chat box in the corner of the screen, which leaves the assistant reading the interface and guessing at what sits behind it.
The backend runs an in-process MCP server, so an agent can call the real operations directly against the real domain model - the same ones behind the human UI.
Asking an agent to investigate a quality problem is then a normal use of the product, working against the domain model directly.
> Why did housing HSG-4471 fail last week? → search_parts("HSG-4471") → get_characteristic_history(char: "Bore Ø12.4", days: 30) 27 sessions, 3 failures - all on 14 Aug, all device CMM-02 → list_device_calibration("CMM-02") calibration expired 11 Aug, renewed 15 Aug → create_ncr_from_failures(session: 8841) NCR-0412 opened, 3 measurements linked The three failures sit inside the window when CMM-02 was out of calibration. NCR-0412 raised for review.
Illustrative transcript. The tool calls shown are real MCP operations.
Both put quality data on a 3D model, so the overlap is a fair question. They cover different halves of the job.
Analyse what happened
Panels you drop into any dashboard, pointed at measurement data you already collect somewhere else. Control charts, capability indices, run rules, Pareto, and characteristics balloonned onto a model for reading trends.
Run the inspection
The system of record. Plans get written and approved here, measurements get captured here, devices and calibration live here, and non-conformances are raised and closed here. The 3D view is there to execute an inspection against real GD&T.
Together they close the loop. The Inspection Manager captures measurements against a plan; the SPC plugins turn that accumulating history into control charts and capability studies. Each works on its own, and a plant running both gets inspection execution and process statistics from one set of numbers.
Everything above is working today. The items below are visible in the product but unfinished. They are listed here so they do not come as a surprise during a trial.
We are looking for a small number of manufacturers to shape this before it opens up, especially teams already working with QIF, AS9102 or a paper NCR process.
English
Polski