Your first request with the OpenAI SDK

In this tutorial you build a small TypeScript script that sends a chat completion through PlainField. By the end, you will have a working client you can reuse for any of PlainField’s 100+ models.

You do not change any provider SDK code — you only change the baseURL and use your PlainField API key. PlainField speaks the OpenAI Chat Completions format, so the OpenAI SDK works unmodified.

Prerequisites

• Node.js  20+
• A PlainField API key (pt_live_...). Create one in the dashboard under Keys → API Keys — see the Quickstart.

1. Create the project

mkdir plainfield-client && cd plainfield-client
npm init -y
npm install openai
npm pkg set type=module

2. Set your API key

Export your PlainField key so the script can read it:

export PLAINFIELD_API_KEY=pt_live_...

3. Write the script

Create index.mjs:

import OpenAI from 'openai'
 
const client = new OpenAI({
  // Point the SDK at the PlainField gateway.
  baseURL: 'https://api.plainfield-ai.com/ai',
  // Your PlainField API key authenticates the request.
  apiKey: process.env.PLAINFIELD_API_KEY
})
 
const response = await client.chat.completions.create({
  model: 'openai/gpt-4o-mini',
  messages: [
    { role: 'system', content: 'You are a helpful assistant.' },
    { role: 'user', content: 'Say hello in one sentence.' }
  ],
  max_tokens: 100
})
 
console.log(response.choices[0].message.content)

4. Run it

node index.mjs

Expected output

Hello! It's great to meet you — how can I help today?

The exact wording varies by model and run; the point is that you get a single line of assistant text back.

What just happened

• The OpenAI SDK sent a standard Chat Completions request to https://api.plainfield-ai.com/ai/chat/completions.
• PlainField read the openai/ prefix on the model field and proxied to OpenAI.
• PlainField logged the request and estimated its cost from the response’s token usage. Open Analytics in the dashboard to see it.

Stream the response

To render tokens as they arrive instead of waiting for the full reply, set stream: true and iterate over the chunks:

const stream = await client.chat.completions.create({
  model: 'openai/gpt-4o-mini',
  messages: [{ role: 'user', content: 'Stream a short poem about the sea.' }],
  stream: true
})
 
for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? '')
}

Streaming works for every model that supports it — PlainField passes the chunks straight through.

Try another provider

Change one string. Switch the model to a different provider:

model: 'anthropic/claude-3-5-sonnet'

See every supported model in Providers and models.

Common errors

Next steps

• Attribute requests to prompts and users, then read their cost: Track prompt costs
• See the full endpoint surface: API reference