Vercel AI SDK connector
nuroute-ai-sdk is the official
Vercel AI SDK provider for NuRoute. generateText, streamText, and
everything else in the AI SDK route through your NuRoute gateway, the same way you'd plug in
@ai-sdk/openai or @ai-sdk/anthropic.
How it works
nuroute-ai-sdk is built on @ai-sdk/openai-compatible,
since the gateway's /v1/chat/completions endpoint already speaks OpenAI's chat completions
format. That means it inherits solid, well-tested request/response/streaming handling for free,
rather than reimplementing the AI SDK's model interface from scratch — this package only adds
NuRoute's own defaults (NUROUTE_API_KEY/NUROUTE_BASE_URL env vars, normalizing a bare gateway
URL to include /v1) and threads NuRoute-specific options through via the AI SDK's standard
providerOptions mechanism (see Forcing a provider below).
Install
npm install nuroute-ai-sdk ai
Quick start
import { createNuRoute } from 'nuroute-ai-sdk';
import { generateText } from 'ai';
const nuroute = createNuRoute({
apiKey: 'aicp-...',
baseURL: 'https://your-nuroute-gateway',
});
const { text } = await generateText({
model: nuroute('auto'), // NuRoute's routing engine picks the best model
prompt: 'Write a haiku about distributed systems.',
});
apiKey/baseURL default to the NUROUTE_API_KEY/NUROUTE_BASE_URL environment variables if
you omit them: createNuRoute() with no arguments works in an environment where those are set.
With nuroute('auto'), Artemis Task picks the model for each request.
Streaming
import { streamText } from 'ai';
const { textStream } = streamText({
model: nuroute('auto'),
prompt: 'Tell me a short story.',
});
for await (const chunk of textStream) {
process.stdout.write(chunk);
}
Forcing a provider
Pin a specific upstream provider via providerOptions, together with a specific model ID to pin
both. With model: 'auto', no router runs: the provider serves the first model it lists.
await generateText({
model: nuroute('claude-sonnet-4-6'),
prompt: 'Hello!',
providerOptions: { nuroute: { provider: 'anthropic' } },
});
Which models can I use?
The same catalog as everywhere else. See Providers & Models for the full list, and the dashboard's Model Catalog tab for the live view (providers your admin has connected platform-wide, plus any you've connected yourself via BYOK).
What's not supported
Only chat completions — no embeddings, image, or completion models, since the gateway doesn't expose those today.