Call Nexus from the OpenAI SDK

You’ll get: your own code calling Nexus with the official OpenAI SDK. Change one address and one key; everything else stays the same.

You need:

• A Nexus API key. Make one in Connections.
• The address of the machine. On the machine itself it’s http://127.0.0.1:3939/v1. Other addresses are on Connections → Gateway access.

Python

from openai import OpenAI

client = OpenAI(
    base_url="http://127.0.0.1:3939/v1",
    api_key="msty_nxs_...",  # your Nexus key, not an OpenAI key
)

reply = client.chat.completions.create(
    model="mlx/Jundot/gemma-4-26B-A4B-it-oQ6e-mtp",
    messages=[{"role": "user", "content": "Say hello in five words."}],
)
print(reply.choices[0].message.content)

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "http://127.0.0.1:3939/v1",
  apiKey: process.env.NEXUS_API_KEY,
});

const stream = await client.chat.completions.create({
  model: "@route/everyday",
  messages: [{ role: "user", content: "Summarize this in one line: ..." }],
  stream: true,
});
for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

What to put in model

Value	What it calls
mlx/Jundot/gemma-4-26B-A4B-it-oQ6e-mtp	One model, by its ID. Copy model id in Models gives it to you.
coding-local	A model alias.
@preset/coder	A preset: a model with saved settings.
@route/everyday	A Smart Route: Nexus picks the model per request.
@pool/qwen-27b	A Smart Balancer pool: Nexus picks the machine.

Streaming, tools and the Responses API work the same way, as long as the model supports them.

Check it worked

List what your key can use:

curl http://127.0.0.1:3939/v1/models -H "Authorization: Bearer $NEXUS_API_KEY"

Then open Activity → Live stream and run your code. Each request shows up as it happens.

If you get 401, check the key. If you get MODEL_NOT_FOUND, check the model name against /v1/models. More in Gateway.