Before you start
- Python 3.10 or later in the notebook's kernel.
- A
memberAPI key, set asCERTONE_API_KEY. If it is not set, the notebook asks for it in a masked prompt and keeps it in memory only.
In the CertOne console, open Developer Hub, then API keys. Name the key, choose the member role and select New key.
- Copy the secret right away: it is shown only once.
- Keys start with
ck_and expire after at most 365 days. - Creating keys requires the
adminrole in your organization. - A
viewerkey can read results but cannot submit scans.
Install and run
Install certone-jupyter, load the extension and scan. The path is the project folder; .. scans the parent of the notebooks folder.
%pip install --quiet certone-jupyter
%load_ext certone_jupyter
sub = %certone_scan .. --asset "Card Fraud Scorer" --tee kata_cc --country CA
sub.verdict, sub.report["by_severity"]Options
| Option | Default | Meaning |
|---|---|---|
path | . | The project folder to scan. |
--asset | folder name | The AI system's name in CertOne. |
--country | ZZ | Where the system operates, for example CA or EU. Note the flag is --country here. |
--tee | amd_sev_snp | amd_sev_snp or kata_cc. |
--fail-on | high | Lowest severity that fails the verdict. |
--api | https://platform.certkor.ai | Or set CERTONE_API_URL. |
Without the magic, call from certone_jupyter import scan in Python.
What you see
- A card titled
CertOne compliance scanwith the verdict in green or red, the counts by severity, the scan id, where it ran and a table of findings with their location. subholds the result:sub.verdict,sub.passed,sub.report,sub.attestation.
Results
The notebook shows the verdict but does not stop anything on its own. To block a model from moving on, use the gate in your pipeline, such as the Azure ML gate.
An error or a scan that does not finish within 30 minutes raises CertOneError.