Autotune CLI preview

autotune is planned as an operator-facing helper for selecting Astra runtime settings from workload intent and host constraints.

Current status

This CLI is not shipped yet.

Use this page as the design contract for future operator guidance, not as a GA command reference.

Planned inputs

• workload type: kubernetes, omni, gateway, or mixed
• host profile: CPU, RAM, disk class, network shape
• durability model: local only vs object-tier or archive backup
• deployment mode: single-node, cluster, or shared multi-tenant backend
• target guardrails: p99 budget, cardinality target, watch fanout expectations

Planned outputs

• recommended Astra profile
• suggested batching and queue-depth settings
• read/LIST path recommendations
• watch dispatch and tenant/auth hints
• warnings when the target host shape is too small for the requested scenario

Until the CLI ships

Use the existing controls directly:

• start from the profile guidance in /operations/profiles-and-auto-governor
• inspect /reference/config/env-vars
• validate every tuning change with the harness scripts in refs/scripts/validation/