Install the OpenAI SDK

The OpenAI SDK is the official client library for the OpenAI API. It's published for Python (openai) and Node (openai) — same API surface. This post installs both and verifies each with a one-line call.

You'll need an API key from platform.openai.com. New accounts get a small credit; pay-as-you-go after that.

Side note: the OpenAI SDK has become the de-facto interface for talking to many non-OpenAI models too. Ollama, LM Studio, vLLM, Together, Groq, and others all expose OpenAI-compatible endpoints. Knowing this SDK is useful even if you don't pay OpenAI directly.

Python

Install via uv inside a project:

# add the openai sdk to your project

uv add openai

For a one-off without a project:

# run a script with the sdk available

uv run --with openai python -c "from openai import OpenAI; print(OpenAI().chat.completions.create(model='gpt-4o-mini', messages=[{'role':'user','content':'hi'}]).choices[0].message.content)"

Legacy projects on pip work fine: pip install openai.

Node

Install via npm inside a project:

# add the openai sdk to your project

npm install openai

Works identically with pnpm and yarn.

Configure the API key

The SDK reads OPENAI_API_KEY from the environment. Don't hardcode it. For dev, use .env (and .gitignore it):

# create .env with your api key

echo 'OPENAI_API_KEY=sk-...' >> .env

For production, use a secret manager. Same advice as any other API key.

Verify (Python)

# verify python sdk with a one-line call

from openai import OpenAI; print(OpenAI().chat.completions.create(model="gpt-4o-mini", messages=[{"role":"user","content":"hi"}]).choices[0].message.content)

Save to verify.py, then uv run python verify.py.

Verify (Node)

// verify node sdk with a one-line call

import OpenAI from "openai"; const r = await new OpenAI().chat.completions.create({ model: "gpt-4o-mini", messages: [{ role: "user", content: "hi" }] }); console.log(r.choices[0].message.content);

Save to verify.mjs, then node verify.mjs.

Pointing the SDK at a different provider

To use the SDK against Ollama, LM Studio, or any OpenAI-compatible endpoint, override base_url (Python) or baseURL (Node) and pass any string as the API key:

# point the openai sdk at a local ollama server

from openai import OpenAI; client = OpenAI(base_url="http://localhost:11434/v1", api_key="ollama")

See Install Ollama or Install LM Studio for the local server side.

Common gotchas

• Wrong env var name: it's OPENAI_API_KEY, exactly. Some old tutorials use OPENAI_KEY — that won't work.
• chat.completions vs responses: OpenAI now has two endpoints. chat.completions is the older, broader one most code uses; responses is newer and tool-use-first. Pick one per project, don't mix.
• Model name drift: OpenAI deprecates models on a schedule. gpt-4o-mini is currently the cheapest reasonable default; check the docs before pinning.
• Rate limits by tier: free credits = tier 1 = low RPM/TPM caps. Tight loops trip them. Use exponential backoff or batch via the Batch API for large jobs (cheaper too).
• Streaming: .stream=True (Python) or .stream: true (Node) returns an iterable of chunks instead of a single response.

With verify.py or verify.mjs printing a reply, you're ready to call OpenAI — or any of the dozens of OpenAI-compatible providers — from the same code.