Use Vercel AI SDK
Use @ai-sdk/openai-compatible for Primex Chat Completions. This explicitly
selects the compatible chat path; other provider packages can default to a
different API. Choose an eligible model in
Vercel AI SDK Setup.
Install and run
npm install ai @ai-sdk/openai-compatible
export PRIMEX_API_KEY=YOUR_API_KEY
export PRIMEX_CHAT_MODEL=YOUR_LIVE_CHAT_MODEL
Save as example.mjs and run node example.mjs on your server:
import { createOpenAICompatible } from '@ai-sdk/openai-compatible';
import { generateText } from 'ai';
const apiKey = process.env.PRIMEX_API_KEY;
const modelId = process.env.PRIMEX_CHAT_MODEL;
if (!apiKey || !modelId) throw new Error('Set PRIMEX_API_KEY and PRIMEX_CHAT_MODEL');
const primex = createOpenAICompatible({
name: 'primex', apiKey,
baseURL: 'https://gateway.primexos.com/v1',
});
const { text } = await generateText({
model: primex(modelId), prompt: 'Hello from Primex', maxRetries: 0,
});
console.log(text);
The answer prints in your terminal. In Next.js, keep this provider in a server route/action; send only the answer to the browser. Never use a public-prefixed environment variable for the key or initialize this client in browser code.
Stream text
In the same script, replace generateText with streamText in the import and
replace the final call/output with:
const result = streamText({
model: primex(modelId), prompt: 'Hello from Primex', maxRetries: 0,
});
for await (const part of result.textStream) process.stdout.write(part);
await result.finishReason;
process.stdout.write('\n');
Confirm a complete final response; partial text is not proof of success. Usage options, structured outputs, tool execution and multi-part tool results must match the selected model/protocol. Do not enable provider capability switches solely because the SDK exposes them. SDK token metadata is not a billing receipt; inspect Usage.
Common fixes
- 401: set the server process environment and restart it after changes.
- 404: use
/v1, the exact offering ID, and the compatible chat provider. - Unsupported option: remove it and review API formats.
- Timeout/stream interruption: inspect Usage before retrying; Errors & limits describes uncertain outcomes.
To disconnect, remove the client configuration and revoke unused keys.
See the official provider reference for additional client settings.