Use this page when something fails. Prefer the section that matches the feature you are using.
| Symptom |
Likely cause |
What to try |
401 on any call |
Bad/missing/revoked API key |
Create a new key; check Authorization: Bearer … |
403 on writes |
Permission, plan feature, or API-key allowlist |
Use a role with ML permissions; try the UI; check plan |
404 on every path |
API URL missing /api/v1 |
Fix LABELTY_API_URL |
| TLS / connection errors |
Wrong host, proxy, or local cert issues |
Ping the host; check corporate proxy settings |
| Symptom |
Likely cause |
What to try |
ModuleNotFoundError: labelty_sdk |
Package not installed |
pip install labelty-sdk in the active env |
| Predictions return 400 about inactive model |
Model not activated |
handle.activate() |
| Download fails after auth works |
Presigned URL / network |
Retry; check clock skew; confirm asset still exists |
client.model_store attribute error |
Outdated snippet |
Use client.org_models / client.marketplace |
| Symptom |
Likely cause |
What to try |
Preview ready_for_training: false |
Too few labeled images / empty selection |
Approve more labels; loosen filters |
Create-run 409 |
Duplicate output name+version |
Change name or output_model_version |
Create-run 403 budget |
Wallet / org budget |
Top up; pick cheaper GPU tier |
Create-run 503 |
SageMaker not configured |
Contact platform ops |
STOPPED_EARLY |
Wallet emptied mid-run |
Top up and resume |
| No epoch metrics |
Worker cannot reach API |
Phases may still move; check ops logs |
| Cannot auto-label after success |
Build not run |
POST …/training/runs/{id}/build |
| Symptom |
Likely cause |
What to try |
503 Lambda / runner not configured |
Ops env missing |
Ask maintainers to configure runner + redeploy |
409 not active / no artifact |
Skipped upload confirm or activate |
Re-run confirm + activate |
404 model name not found |
Wrong UUID in deploy_config.models |
Use an active registered model id |
| Slow first call (~tens of seconds) |
Cold start |
Use 120s client timeout; warm with a ping image |
429 |
Another run in progress on this process |
Retry after the in-flight call finishes |
route_deliveries error |
DatasetOutput ingest failed |
Predictions may still be valid; check ingest config with ops |
| Symptom |
Likely cause |
What to try |
| Cannot open / billing error |
Wallet below minimum or plan flag off |
Top up wallet; verify plan |
| Presigned URL expired |
Waited too long |
Click Open JupyterLab again |
| Env vars missing in kernel |
Lifecycle did not run / stale app |
Stop → Open; verify CloudWatch lifecycle logs |
import labelty_sdk fails |
Install from wrong index |
Manual pip install from the team’s index |
| GPU instance but CPU-only libs |
Image not switched for GPU |
Confirm with ops; see open questions |
- Capture the HTTP status, request path, and error payload (redact secrets).
- Note the project id, run/workflow id, and approximate timestamp.
- Check Open questions for known documentation gaps.
- Escalate to your Labelty admin or platform operator with that context.
