Integrations
Model Naming
How to name the model in your request so the gateway routes it to the right provider.
The rule
- A model name without a slash (for example
gpt-4o) uses the gateway's default mapping. - A model name that contains a slash requires a provider prefix. Write it as
<provider>/<model>. - The provider prefix is used for routing and is stripped before the request goes upstream: the provider only sees the model name.
Examples
| Model name you send | Routed to | Sent upstream as |
|---|---|---|
openai/gpt-4o | OpenAI | gpt-4o |
anthropic/claude-3-5-sonnet-20240620 | Anthropic | claude-3-5-sonnet-20240620 |
gemini/gemini-3.5-flash | Google Gemini | gemini-3.5-flash |
groq/llama-3.1-8b-instant | Groq | llama-3.1-8b-instant |
groq/openai/gpt-oss-20b | Groq | openai/gpt-oss-20b |
The last example matters: the upstream model id itself contains a slash (
openai/gpt-oss-20b), so you prefix it with the provider you are calling through (groq/) and the gateway strips only that first segment.Use the same name in scopes
Model scopes use the model name, e.g. gateway:model:gpt-4o. See Scopes.
Provider keys
Whichever provider the prefix selects, the gateway looks up your key for that provider in the Vault. If none is stored the request fails. Add it in the dashboard Vault first.