fleet-updating-pricing

To update model pricing data for AI cost estimation, trigger the pricing refresh to fetch current vendor prices and write them into scripts/fleet/constants/model-pricing.json with today's date.

The pricing snapshot refresh works by orchestrating four phases: reading the current snapshot, fetching latest vendor prices, applying new prices to model-pricing.json, and restamping the snapshot marker in docs to ensure cost accuracy.

Yes, you need WebFetch to retrieve current vendor prices during the refresh cycle. The node-based scripts then use the fetched data to write updated pricing and restamp the snapshot marker.

The best way to automate weekly pricing cycles is using the on-demand /update-pricing endpoint, which fits into the weekly updating umbrella to maintain per-model pricing accuracy for the fleet's AI cost estimator.

You should trigger an on-demand pricing refresh when vendor prices move unexpectedly, ensuring your AI cost estimator stays accurate without waiting for the standard weekly pricing cycle to execute.