1# Dynamic Routing
2 
3Configure complex routing in dashboard without code changes. Use route names instead of model names.
4 
5## Usage
6 
7```typescript
8const response = await client.chat.completions.create({
9  model: 'dynamic/smart-chat', // Route name from dashboard
10  messages: [{ role: 'user', content: 'Hello!' }]
11});
12```
13 
14## Node Types
15 
16| Node | Purpose | Use Case |
17|------|---------|----------|
18| **Conditional** | Branch on metadata | Paid vs free users, geo routing |
19| **Percentage** | A/B split traffic | Model testing, gradual rollouts |
20| **Rate Limit** | Enforce quotas | Per-user/team limits |
21| **Budget Limit** | Cost quotas | Per-user spending caps |
22| **Model** | Call provider | Final destination |
23 
24## Metadata
25 
26Pass via header (max 5 entries, flat only):
27```typescript
28headers: {
29  'cf-aig-metadata': JSON.stringify({
30    userId: 'user-123',
31    tier: 'pro',
32    region: 'us-east'
33  })
34}
35```
36 
37## Common Patterns
38 
39**Multi-model fallback:**
40```
41Start → GPT-4 → On error: Claude → On error: Llama
42```
43 
44**Tiered access:**
45```
46Conditional: tier == 'enterprise' → GPT-4 (no limit)
47Conditional: tier == 'pro' → Rate Limit 1000/hr → GPT-4o
48Conditional: tier == 'free' → Rate Limit 10/hr → GPT-4o-mini
49```
50 
51**Gradual rollout:**
52```
53Percentage: 10% → New model, 90% → Old model
54```
55 
56**Cost-based fallback:**
57```
58Budget Limit: $100/day per teamId
59  < 80%: GPT-4
60  >= 80%: GPT-4o-mini
61  >= 100%: Error
62```
63 
64## Version Management
65 
66- Save changes as new version
67- Test with `model: 'dynamic/route@v2'`
68- Roll back by deploying previous version
69 
70## Monitoring
71 
72Dashboard → Gateway → Dynamic Routes:
73- Request count per path
74- Success/error rates
75- Latency/cost by path
76 
77## Limitations
78 
79- Max 5 metadata entries
80- Values: string/number/boolean/null only
81- No nested objects
82- Route names: alphanumeric + hyphens
83