Automatic Routing
Understand exact and wildcard routing policies, priority ordering, and provider fallback resolution.
Zorveus dynamically evaluates attached provider credentials to route each inference request to the most cost-effective and reliable upstream provider.
Instead of writing provider-specific fallback code or managing multiple API keys in your application, Zorveus handles model routing automatically.
How Provider Resolution Works
When an inference request arrives at https://api.zorveus.com/v1/chat/completions, Zorveus evaluates the requested model identifier against your organization's attached provider credentials.
Request Model: "openai/gpt-4.1-mini"
│
├─► 1. Check Active BYOK Credentials matching "openai/*" or "openai/gpt-4.1-mini"
├─► 2. Sort by Priority (Higher numeric priority selected first)
├─► 3. Health Check Verification (Exclude failing or rate-limited credentials)
└─► 4. Fallback to Zorveus Managed Gateway (If balance is available)Model Matching Patterns
Provider credentials can specify exact model matches or wildcard patterns:
| Scope Pattern | Description | Example Matches |
|---|---|---|
* | Global match across all providers and models. | openai/gpt-4.1-mini, anthropic/claude-3-7-sonnet |
openai/* | Matches all models under the OpenAI namespace. | openai/gpt-4.1-mini, openai/o3-mini |
anthropic/* | Matches all models under the Anthropic namespace. | anthropic/claude-3-7-sonnet, anthropic/claude-3-5-haiku |
openai/gpt-4.1-mini | Restricts key usage strictly to a single model. | openai/gpt-4.1-mini |
Priority and Resolution Rules
- Active Status Check. Credentials toggled off or flagged with consecutive upstream API errors are excluded from routing.
- Priority Score. If multiple credentials match the requested model, Zorveus selects the credential with the highest priority integer (
priority: 30takes precedence overpriority: 10). - Round-Robin Equal Priority. When credentials share identical priority values, Zorveus balances traffic evenly between them.
- Key Cooldown. If a credential is rate-limited or fails health checks, Zorveus temporarily excludes it from routing until it recovers. In that scenario, the next highest priority credential is selected.
- Zorveus Managed Fallback. If no custom provider credential matches or if all BYOK routes fail, Zorveus routes the request through the Zorveus Managed Gateway, charging your organization wallet balance directly.
Configure priority in the dashboard
Open Provider Credentials in the dashboard and configure the credential's routing priority. Application code does not update provider credentials through an API.