Provider Routing

Share

KiteRouter exposes an OpenAI-compatible API to customers. Admins can map one customer-facing model to multiple upstream provider models, ordered by priority.

How model routing works

  1. The client sends model: "gpt-4o-mini" to /api/v1/chat/completions.
  2. KiteRouter authenticates the key, checks credits, and loads active provider mappings for that model.
  3. KiteRouter rewrites the public model ID to the selected upstream model ID.
  4. The request body is forwarded to the provider's OpenAI-compatible endpoint.
  5. If the first provider fails, KiteRouter can try the next active mapping.

Public model vs upstream model

NameExampleWho sees it
Public model IDgpt-4o-miniYour customers and API users
Upstream model IDopenai/gpt-4o-mini or provider-specific IDAdmin/provider mapping only
Provider base URLhttps://api.openai.com/v1Admin/provider configuration

Warning: Send OpenAI-compatible request bodies. Provider-specific differences belong in admin provider configuration and upstream compatibility.

Ready to start building?

Get your API key and make your first request in under a minute.