SDK install and first call

Загрузка…

SDK install and first call — OpenAI API from A to Z

Install

npm install openai

Create a key at platform.openai.com/api-keys and put it in the OPENAI_API_KEY environment variable. Never hardcode the key in your repo.

Minimal TypeScript call

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY, // picked up automatically, but be explicit
});

const completion = await client.chat.completions.create({
  model: "gpt-4o",
  messages: [
    { role: "system", content: "You are a terse technical assistant." },
    { role: "user", content: "Explain closures in JS in 3 sentences." },
  ],
});

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

What's in the response

The chat.completion response object contains:

id — identifier (like chatcmpl-...), useful for OpenAI support debugging.
choices — an array of completions (one by default). Each choices[i].message has role and content.
choices[i].finish_reason — "stop" (model finished on its own), "length" (hit max_tokens), "tool_calls" (model wants to call a function), "content_filter".
usage — prompt_tokens, completion_tokens, total_tokens for cost accounting.
model — the exact model that actually ran (e.g. gpt-4o-2024-08-06).

Same in Python

from openai import OpenAI

client = OpenAI()  # reads OPENAI_API_KEY from env

completion = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        {"role": "system", "content": "You are a terse assistant."},
        {"role": "user", "content": "Hello"},
    ],
)
print(completion.choices[0].message.content)
print(completion.usage.total_tokens)

Limiting response length

max_tokens (for o-models: max_completion_tokens) caps only the response length, not the context. If the answer was cut off, finish_reason will be "length". Without it, the model decides when to stop and sets finish_reason: "stop".

Сообщить об ошибке

Квиз — 3 вопроса

Install

Minimal TypeScript call

What's in the response

Same in Python

Limiting response length