Gateways
Planned
The gateway adapters are designed but not yet built.
Organizations that already route LLM traffic through an API gateway get Lumen as a plugin at that choke point — no per-machine deployment, full inline enforcement for every application behind the gateway.
Design
Each adapter is a thin plugin in the gateway's native extension language that calls the Lumen inline verdict API on the request and response bodies:
| Gateway | Plugin form |
|---|---|
| Kong | Lua plugin |
| LiteLLM | Python callbacks (async_pre_call_hook / async_post_call_*) |
| Portkey | TypeScript/JS plugin |
- Requests are evaluated before they are forwarded upstream: Redact rewrites the body, Block returns the provider's own error shape so client SDKs fail cleanly.
- Responses — streaming included — are evaluated before delivery, through the same hold-and-release window as every other inline collector.
- Registration works like every collector: the plugin carries a tenant-scoped API key and a policy assignment from the console.
When to choose a gateway collector
- You already operate Kong / LiteLLM / Portkey in front of your model traffic.
- You want one enforcement point for many applications, rather than instrumenting each with the SDK.
- The traffic you care about is server-side. Developer laptops and browsers still need the host-local collectors.