DEV Community: routerbasecom

Point the OpenAI SDK at RouterBase in a few minutes

routerbasecom — Tue, 30 Jun 2026 17:17:47 +0000

If your application already uses OpenAI-style chat-completions calls, the lowest-friction way to test another routing layer is to keep the client code familiar and change only the base URL, API key, and model id.

RouterBase is designed for that kind of experiment. It exposes an OpenAI-compatible endpoint at https://routerbase.com/v1, so existing SDK setup can stay close to what developers already know.

Environment variables

ROUTERBASE_API_KEY=your_routerbase_key
ROUTERBASE_BASE_URL=https://routerbase.com/v1
ROUTERBASE_MODEL=google/gemini-2.5-flash

Keeping the model in configuration makes it easier to compare providers later without editing the application path that calls the model.

JavaScript SDK setup

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.ROUTERBASE_API_KEY,
  baseURL: process.env.ROUTERBASE_BASE_URL || "https://routerbase.com/v1"
});

const completion = await client.chat.completions.create({
  model: process.env.ROUTERBASE_MODEL || "google/gemini-2.5-flash",
  messages: [
    {
      role: "user",
      content: "Write a one-paragraph changelog entry for a developer tool."
    }
  ]
});

console.log(completion.choices[0]?.message?.content);

What to verify first

Before using the setup in a production workflow, run a short checklist:

The request succeeds with the RouterBase base URL.
The model id is read from configuration, not hard-coded deep in the codebase.
Errors are logged with status code and request context.
The team has a small comparison prompt set for latency and output quality.

Useful links

RouterBase
RouterBase docs: https://docs.routerbase.com/
npm quickstart package: https://www.npmjs.com/package/routerbase-quickstart

Use RouterBase as an OpenAI-compatible API gateway

routerbasecom — Tue, 30 Jun 2026 17:10:01 +0000

Many AI applications begin with one provider and then need more flexibility: fallback models, model experiments, or different models for different product workflows.

RouterBase gives developers an OpenAI-compatible API shape at https://routerbase.com/v1, so the first experiment can stay small: change the base URL, set a RouterBase API key, and choose a model id.

Minimal curl request

curl https://routerbase.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERBASE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/gemini-2.5-flash",
    "messages": [
      { "role": "user", "content": "Explain RouterBase in one sentence." }
    ]
  }'

Node.js example

const response = await fetch("https://routerbase.com/v1/chat/completions", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.ROUTERBASE_API_KEY}`,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: process.env.ROUTERBASE_MODEL || "google/gemini-2.5-flash",
    messages: [
      { role: "user", content: "Draft a short product update." }
    ]
  })
});

if (!response.ok) {
  throw new Error(`RouterBase request failed: ${response.status}`);
}

console.log(await response.json());

Practical rollout advice

Start with a low-risk workflow, keep the model id configurable, and compare output quality, latency, cost, and error rate before routing more traffic through the new setup.

Good first workflows:

Internal release note drafts
Support ticket summaries
Model comparison prompts
Non-critical classification tasks

Useful links

RouterBase
RouterBase docs: https://docs.routerbase.com/
npm quickstart package: https://www.npmjs.com/package/routerbase-quickstart