Resilience & Fallback

Keep chatting even when providers fail.

Resilience Settings Overview

Access via Settings → Resilience.

Why Resilience Matters

AI APIs can fail:

• Rate limits when you hit usage caps
• Network errors during connectivity issues
• Auth failures when keys expire
• Outages when providers go down

Chorum’s fallback system automatically switches to alternatives so you’re never stuck.

Settings Explained

Enable Automatic Fallback

Toggle: On/Off

When enabled:

• Primary provider fails → automatically tries next configured provider
• Happens silently in the background
• You see the response, not the failure

Default Fallback Provider

Dropdown selection

If something breaks, which provider should Chorum try first?

This is your “known good” choice — typically your most reliable or cheapest provider.

Example: Set to Google (Gemini) as a reliable, cost-effective backup.

Offline Fallback (Local Model)

Text field — Enter a local model name

If all cloud providers fail (internet outage?), Chorum will try a local model via Ollama.

Common options:

• llama3 — Good balance of speed and quality
• mistral — Fast, lightweight
• phi3 — Microsoft’s efficient model
• codellama — For code-heavy work
• gemma2 — Google’s local model

Leave empty to skip local fallback.

How Fallback Works

The decision chain:

1. Primary provider fails
   ↓
2. Try default fallback provider
   ↓
3. Try other configured cloud providers
   ↓
4. Try local model (if configured)
   ↓
5. Show error (only if all fail)

What Triggers Fallback

Best Practices

1. Configure 2+ cloud providers — Diversity = reliability
2. Set a default fallback — Your “safe” choice
3. Add a local model — True offline capability
4. Test your fallback — Disable a provider and verify switching works

Example Setup

This setup means you’d need cloud + local to all fail before seeing an error.

→ Back to: Settings Overview