LLM gateway

An LLM gateway (AI API gateway) sits between your app and model providers: one API, one key, routing, and cost control. Geek Hub is that gateway — OpenAI-compatible, built for Mexico and LATAM teams.

What an LLM gateway is

Without a gateway, every provider (OpenAI, Anthropic, Google, …) means another SDK, invoice, and policy. With a gateway, your code speaks one protocol — usually OpenAI-compatible — while the server picks models, enforces limits, and meters cost.

You want it when you need to swap models without redeploying, fail over on outages, budget by key, and audit usage in one place. It is not a chat UI — it is API infrastructure.

What Geek Hub does as a gateway

One OpenAI-compatible base URLhttps://api.geekhub.mx/v1 — chat, images, video, TTS, embeddings on the same client.
Routing and failoverPass a model list or model:"auto"; if the first fails, the gateway retries.
MX billing and compliancePay-as-you-go balance, CFDI 4.0 on top-ups, per-request cost visibility.
Guardrails and ZDROrg/key policies and rejection when you request zero retention on an unverified model.

Minimal example

Point the OpenAI SDK at the gateway. The model field picks the maker.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.geekhub.mx/v1",
  apiKey: process.env.GEEKHUB_API_KEY,
});

const res = await client.chat.completions.create({
  model: "google/gemini-2.5-flash",
  messages: [{ role: "user", content: "Summarize in one line" }],
});

Gateway vs going direct

Need	Direct (N APIs)	Geek Hub (1 API)
Swap models	Code + keys per provider	Change the model string
Maker outage	Your feature dies	Fallback to the next option
Mexican invoicing	N USD invoices	Top-ups + central CFDI
Budget per environment	Manual / missing	Per API key limit

FAQ · LLM gateway

Is an LLM gateway just a proxy?A pure proxy forwards. A gateway adds auth, metering, routing, policy, and often a unified catalog. Geek Hub is the latter.
How much latency does it add?Typical hop is a few ms. In return you avoid app-side retries and saturated providers.
Where is pricing?See GET /v1/models and /en/models. Pay-as-you-go on top of provider price.

Stand up your gateway in minutes

Get started