Skip to content

Over-the-Air Updates

modelBridge updates itself continuously in the background. Error messages, pricing formulas, and operational controls refresh automatically — so the plugin you’re using today is more accurate than it was yesterday, even without installing a new version.

When a new error pattern appears on fal.ai — a new validation type, a changed error message, or a new model-specific failure — modelBridge can ship updated error copy within an hour. You see a clear, actionable message instead of a generic fallback, without waiting for a plugin release.

When fal.ai changes a model’s pricing — new resolution tiers, audio surcharges, or rate adjustments — modelBridge’s curated pricing supplements can be updated remotely. Your cost estimates stay accurate between plugin versions.

New features can be rolled out gradually or disabled instantly via remote feature flags. If a feature causes issues, it can be turned off for all users within an hour — no plugin update, no manual action required.

During fal.ai outages or planned maintenance, modelBridge can display a service notice banner automatically. The notice appears and disappears based on the current status — you’re warned before wasted API credits, and the warning goes away when the issue resolves.

At plugin startup, modelBridge fetches a small configuration file from a known URL. This file contains versioned error copy, pricing data, feature flags, and any active incident notices. The fetch has a 3-second timeout and uses a stale-while-revalidate pattern:

  1. Serve cached data immediately — the plugin starts with whatever data it has from the last session
  2. Fetch fresh data in the background — if newer data is available, it’s applied silently
  3. Fall back gracefully — if the fetch fails (offline, timeout, server error), cached data is used without interruption

The configuration refreshes approximately every hour during active use. No user action is required.

Remote updates are limited to copy, configuration, and metadata. They cannot:

  • Change plugin behavior or logic (that requires a plugin update)
  • Access your data, API keys, or project files
  • Install or remove models
  • Modify your settings or preferences
  • Execute code on your machine

The configuration fetch sends only the plugin version and platform to determine compatibility. No personal data, usage history, or project information is included in the request.