Models (sidebar: INSIGHTS → Models, /models) manages production ML assets for Command-tier organizations. GIBE (Gravitre Intelligent Business Engine) powers the Built-in models catalog at /models/built-in — org-scoped models trained on verified signals without a manual training job.
Open the registry
The overview shows totals for registered, deployed, and in-training models plus connected data Sources available for training.
Pick a stack layer template (generative, classical, forecasting, anomaly, MLOps) to pre-fill a create dialog.
Browse GIBE built-in models
- Open INSIGHTS → Built-in models (
/models/built-in) - Review available GIBE models — failure predictors, rankers, duration forecasts, and similar
- Check signal coverage and last evaluation date before relying on a built-in model in production
- Deploy or attach built-in models to workflows when status is ready
Built-in models update as your org accumulates runs, chat turns, and promoted memories. See GIBE — Learning for evaluation and quality tools.
Register a model
- Click New model or apply a layer template
- Set name, model type (classifier, fine-tuned LLM, anomaly detector, forecaster)
- Choose base model and optional dataset / task type
- Save — status starts as draft
Model types map to inference and training behavior documented in OpenAPI.
Train
Start training from the model detail page or via API:
curl -s -X POST "https://gravitre.app/api/ml/train/start" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model_id": "MODEL_ID"}'
Monitor status transitions: training → validating → ready (or failed).
Deploy
When a model is ready, deploy to make it callable in production:
curl -s -X POST "https://gravitre.app/api/ml/models/MODEL_ID/deploy" \
-H "Authorization: Bearer YOUR_API_KEY"
Deployed models show a deployed badge in the registry list.
Run inference
curl -s -X POST "https://gravitre.app/api/ml/models/MODEL_ID/predict" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"features": {}}'
Request body shape depends on model type — inspect the model detail page and OpenAPI schema.
Connect data Sources
Generative and tabular models may require Sources or Connectors marked as connected in the registry UI. Connect data before training if the template lists requires_connection.
List and filter
Use list filters by status and model type. Open a model row for version metadata, linked workflows, and deployment history.
API summary
| Action | Method | Path |
|---|---|---|
| List models | GET | /api/ml/models |
| Get model | GET | /api/ml/models/{id} |
| Create | POST | /api/ml/models |
| Deploy | POST | /api/ml/models/{id}/deploy |
| Predict | POST | /api/ml/models/{id}/predict |
| Train | POST | /api/ml/train/start |
All /api/ml/* routes require Command plan entitlements.
Former name
Docs and UI previously said Model Registry or AI Models. User-facing label is now Models; GIBE built-in catalog lives at /models/built-in. API paths remain /api/ml.