Documentation
Availability and known limitations
On this page
This page separates implementation from proof of customer-ready behavior.
| Capability | Status in the reviewed preview | Practical limit |
|---|---|---|
| Public Lab image workflow | Implemented | Browser development output; no camera/service connection or production acceptance |
| Synthetic sample and mock HMI | Implemented simulation | UI/data demonstration only |
| Browser filters, value nodes and blob location | Implemented; software conformance evidence | Independent application-quality and target-browser qualification still required |
| Windows camera adapters and job browsing | Implemented with historical device-specific evidence | Only the approved package/model/firmware/site combination may be relied on |
| Camera writes, baseline and recovery records | Implemented controls | Unknown outcomes must be reconciled; complete production recovery is not established |
| Four/eight-camera concurrency | Software/simulation evidence | Multi-device real throughput and fault isolation still need hardware acceptance |
| All camera models / line scan / L38 3D | Incomplete or unqualified | No blanket model claim; point-cloud display/export not implemented; D900 excluded from Web HMI |
| Browser OCR/reference/custom-code candidates | Separate candidate builds | Excluded from the default public Lab; no production-quality guarantee |
| Private account and own-Codex relay | Research/candidate implementation | Not public Lab features; real identity/provider and complete journey/release validation pending |
| Dataset/candidate/release flow | Implemented engineering code | Does not prove installer or line acceptance; physical-item and negative/holdout coverage matter |
| Clean Windows install, downgrade and long run | Acceptance pending/incomplete | Do not infer from source build or historical targeted tests |
| Signed installer | Not established; the reviewed security note says unsigned | Confirm exact supplied package status before customer distribution |
| Documentation languages | English and eight translated editions | Translation does not establish product language support or hardware acceptance. |
Quality and performance
A software test can establish a contract or algorithmic equivalence on its fixtures. It cannot establish recognition accuracy on unseen physical products. Historical OCR candidate quality gates failed; they must not be presented as a qualified OCR feature. Historical two-item barcode development measurements lack an independent holdout/negative set and cannot be generalized to production.
Elapsed browser/PC calculation time excludes some combination of acquisition, transport, UI, startup or PLC handling. Use the timing boundary attached to each record; establish a site cycle-time budget and measure the entire intended path on target hardware.
The dataset store currently limits a dataset to 2,000 images. Repeating a physical item does not provide independent additional acceptance samples. If your statistical qualification requires larger independent sets, agree a supported process before using the release gate.
Confirm for your package
These items are confirmed per package and per site rather than stated on this page: the commercial edition and final installer version, the Windows support matrix, the field labels for each accepted camera job, the Browser Lab privacy and consent wording, and your support route. Ask DeYiVision through the contact page.