vmanomaly-sizing
Installation
SKILL.md
Deployment sizing
Currently experimental as of v1.31.0. Use the connected server's deployment-sizing API (v1.31.0+) or the corresponding MCP tools. Estimate vmanomaly resources without changing its configuration. No exact query, data fetch or autotune is needed when the model and workload shape are already known.
Choose the unknown
| User knows | Estimate | MCP tool | HTTP route |
|---|---|---|---|
| Model, series count, inference cadence | Required resources | vmanomaly_estimate_deployment_resources |
POST /api/v1/deployment-sizing/estimate |
| Model, CPUs, cadence; optionally RAM | Active models served per interval | vmanomaly_estimate_inference_capacity |
POST /api/v1/deployment-sizing/throughput |
| Wants supported model configurations | Available sizing profiles | vmanomaly_get_deployment_sizing_profiles |
GET /api/v1/deployment-sizing/profiles |
For reverse sizing, cardinality is the output: do not ask for a series count. Clarify missing history, sampling or model parameters when needed, or disclose defaults. Preserve an explicitly supplied version. Check model schemas or available profiles before supplying required parameters. Unsupported versions/configurations are not permission to substitute another model or seasonality.