Status: Planned

​
Drift Detection — How it would work

​
Problem & goals

  • Detect when real-world resources drift from the last applied configuration.
  • Provide safe, read-only detection with clear surfacing in CLI/UI.

​
User journeys

  • Configure drift checks per unit or prefix (schedule or on-demand).
  • View drift status in taco unit status and UI; inspect drift details.

​
High-level design

  • Read-only checks that fetch remote resource state via providers.
  • Store minimal drift metadata alongside unit records; do not block writes.
  • Respect locks, rate limits, and provider quotas.

​
Shapes (provisional)

  • API: POST /v1/drift/check (on-demand), GET /v1/units/:id/drift (status).
  • Status values: unknown | clean | drifted; timestamps; summary.

​
Open questions

  • Scheduling window; integration with remote runs for remediation.
  • Scope/precision for multi-provider environments.
Note: Subject to change; no timelines implied.