mirror of
https://github.com/ksyasuda/dotfiles.git
synced 2026-03-20 06:11:27 -07:00
1.8 KiB
1.8 KiB
Dynamic Routing
Configure complex routing in dashboard without code changes. Use route names instead of model names.
Usage
const response = await client.chat.completions.create({
model: 'dynamic/smart-chat', // Route name from dashboard
messages: [{ role: 'user', content: 'Hello!' }]
});
Node Types
| Node | Purpose | Use Case |
|---|---|---|
| Conditional | Branch on metadata | Paid vs free users, geo routing |
| Percentage | A/B split traffic | Model testing, gradual rollouts |
| Rate Limit | Enforce quotas | Per-user/team limits |
| Budget Limit | Cost quotas | Per-user spending caps |
| Model | Call provider | Final destination |
Metadata
Pass via header (max 5 entries, flat only):
headers: {
'cf-aig-metadata': JSON.stringify({
userId: 'user-123',
tier: 'pro',
region: 'us-east'
})
}
Common Patterns
Multi-model fallback:
Start → GPT-4 → On error: Claude → On error: Llama
Tiered access:
Conditional: tier == 'enterprise' → GPT-4 (no limit)
Conditional: tier == 'pro' → Rate Limit 1000/hr → GPT-4o
Conditional: tier == 'free' → Rate Limit 10/hr → GPT-4o-mini
Gradual rollout:
Percentage: 10% → New model, 90% → Old model
Cost-based fallback:
Budget Limit: $100/day per teamId
< 80%: GPT-4
>= 80%: GPT-4o-mini
>= 100%: Error
Version Management
- Save changes as new version
- Test with
model: 'dynamic/route@v2' - Roll back by deploying previous version
Monitoring
Dashboard → Gateway → Dynamic Routes:
- Request count per path
- Success/error rates
- Latency/cost by path
Limitations
- Max 5 metadata entries
- Values: string/number/boolean/null only
- No nested objects
- Route names: alphanumeric + hyphens