Core Concepts
How the Gateway Works
Plurence sits between your application and AI providers, adding routing, policy enforcement, and observability without changing your code.
Request flow
Every request travels through three stages:
Authenticate
The gateway validates your API key against a project. The project determines which policies and quotas apply.
Policy check
Rate limits, quota ceilings, and content policies are evaluated. Requests that exceed limits are rejected before reaching the upstream provider.
Route & proxy
The request is translated to the target provider's format and forwarded. The response is translated back to the OpenAI format and returned to your client.
Two endpoints
Plurence exposes two distinct services:
| Service | URL | Auth |
|---|---|---|
| Gateway | gateway.plurence.com | Project API key |
| Management API | data.plurence.com | User JWT |
Your application code only ever talks to the gateway. The management API is used by the dashboard and for programmatic account management.
OpenAI compatibility
The gateway implements the OpenAI REST API surface — /v1/chat/completions,
/v1/embeddings, and streaming via Server-Sent Events.
Any client that works with OpenAI works with Plurence without code changes.
Observability
Every gateway request produces an audit event and a usage record. Audit events capture auth outcomes, policy decisions, and errors. Usage records track token counts and latency for billing and monitoring. Both are queryable from the Audit log and Usage panels.