Documentation

Proximarun Documentation

Reference and guides for the Proximarun routing engine. The quickstart covers the common case in 15 minutes. The concepts guide covers the mental model you need to configure routing policies and debug production behavior.

Quickstart

Install the SDK, connect your node pool, and run your first routed inference call in under 15 minutes.

API Reference

REST endpoint reference for POST /infer, GET /nodes, POST /nodes, DELETE /nodes/{id}, GET /telemetry, and GET /status. Includes request schemas, response shapes, and curl examples.

Concepts

The six building blocks of the routing model: node pool, routing policy, SLA targets (latencyBudgetMs, costCeilingUsd), traffic pattern learning, health check cycle, and telemetry schema. Start here if routing behavior is surprising.

Engineering Blog

Technical posts on inference routing, GPU cost engineering, and node placement decisions from the team building Proximarun. Not a marketing blog.

Get started

Three steps to your first routed call.

1

Install the SDK

npm install @proximarun/client

2

Initialize the client

import { ProximaClient } from '@proximarun/client';

const client = new ProximaClient({
  apiKey: process.env.PROXIMA_KEY
});

3

Make your first routed call

const result = await client.infer({
  prompt: messages,
  latencyBudgetMs: 400,
  costCeilingUsd: 0.002
});

That is all. Proximarun selects the node that fits your declared budget. The response is returned unchanged. Read the full quickstart for node pool setup and telemetry verification.