The modelscan agent enumerates local model artefacts on a customer host (public-registry model caches, GGUF blobs, ONNX files, custom directories) and reports back signed-SHA-256 hashes plus structural verdicts. Each modelscan_run is one execution of that agent against a base directory.
GET /api/v2/modelscan-runs
curl -sS "https://penaxtra.com/api/v2/modelscan-runs" \
-H "Authorization: Bearer $TOKEN"Required scope: modelscan:read.
Response
{
"data": [
{
"id": "<uuid>",
"agent_id": "<uuid>",
"hostname": "ml-builder-02",
"os": "linux",
"arch": "amd64",
"agent_version": "0.6.1",
"started_at": "2026-05-22T10:14:00Z",
"completed_at": "2026-05-22T10:22:11Z",
"status": "completed",
"files_scanned": 184,
"bytes_scanned": 73120442889,
"findings_count": 2,
"critical_count": 1,
"high_count": 1,
"duration_ms": 491233
}
]
}Fields
| Field | Notes |
|---|---|
id | Run UUID. |
agent_id | The enrolled modelscan agent that produced this run. |
hostname / os / arch | Agent reporting metadata at run start. |
agent_version | Self-reported agent build. |
started_at / completed_at | Wall-clock timestamps. |
status | running, completed, failed. |
files_scanned / bytes_scanned | Coverage size; useful for capacity planning. |
findings_count / critical_count / high_count | Counts by severity. |
duration_ms | End-to-end runtime in milliseconds. |
Notes
- Newest run first; result set capped at 100 rows.
- Modelscan findings (per-file verdicts) are addressable via the standard findings endpoint with
target_kind=model_artifact. - Triggering a run from the API is not exposed; runs are launched by the on-host agent itself on a schedule or via the console-managed CLI.
Related
Last reviewed: 2026-06-13. Reviewed by: Engineering. Content type: Developer documentation. Reach the maintainers: [email protected] .