ADR 0002: Cost Estimation (Pre-Flight Spend Alerts)

Status

Proposed

Context

As part of the rune-ui initiative, the rune core must be extended to provide cost estimation logic for all benchmarks before they are executed. This ensures that both the CLI and UI can provide consistent "Pre-Flight Spend Alerts."

Decision

1. Unified Logic: Centralize all cost calculation logic in the rune_bench core.
2. API Endpoint: Implement POST /v1/estimates in rune_bench/api_server.py.
3. Estimation Drivers:
   • Cloud: Vast.ai (current market rates), stubs for AWS/GCP/Azure.
   • Local: Energy consumption (TDP * rate * duration) and hardware amortization.

Consequences

• CLI and UI can provide spend alerts before job submission.
• Consistent cost modeling across all interfaces.
• Requires maintenance of market rate queries and energy models.