Quick Start

The Inference API is OpenAI-compatible. Point an OpenAI SDK at https://api.inference.nebul.io/v1, pass your API key, and call any model in the catalog. This page takes you from a fresh key to a first response, in the language you already use.

Prerequisites

• A Nebul account with API access. Sign up at Nebul AI Studio.
• An API key created in AI Studio. Keys start with sk-.
• Python 3.9 or higher for the Python example; every other tab needs only its OpenAI SDK or an HTTP client.

Export your key in your terminal before running any example on this page:

export NEBUL_API_KEY="sk-your-key"

Make your first request

Pick a language. Each tab installs the client and sends one chat request to zai-org/GLM-5.3:

pip install openai

import os

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["NEBUL_API_KEY"],
    base_url="https://api.inference.nebul.io/v1",
)

response = client.chat.completions.create(
    model="zai-org/GLM-5.3",
    messages=[{"role": "user", "content": "Say OK"}],
)

print(response.choices[0].message.content)

You should see the model's answer to your message. If you get a 401, the key is wrong or not exported; a 404 means the model isn't enabled for your project, so swap in one from the list below.

The snippets call zai-org/GLM-5.3. Replace it with any model ID from GET /v1/models or the Model Catalog; IDs look like zai-org/GLM-5.3 and openai/gpt-oss-120b. The SDKs page covers the same setup in more environments, including plain fetch and streaming.

List available models

Every key is scoped to a project, so the list shows what that project can call:

import os

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["NEBUL_API_KEY"],
    base_url="https://api.inference.nebul.io/v1",
)

for model in client.models.list():
    print(model.id)

One model ID prints per line. Fine-tuned adapters appear too, with IDs starting ft:. To browse models with pricing and context windows instead, use the Model Catalog.

Endpoints

Error handling

Errors come in two shapes. Middleware errors (authentication, routing, limits, budgets) return {"detail": {"message", "error_type"}}. Errors raised inside the model backend use the OpenAI shape {"error": {"message", "type", "param", "code"}}:

{
  "detail": {
    "message": "Invalid API key provided",
    "error_type": "unauthorized_error"
  }
}

Every status code, response body, and retry rule is covered in Errors.

Next steps

• Examples: streaming, embeddings, vision, tools, and reasoning
• SDKs: the same setup in Python, TypeScript, JavaScript, Go, Rust, Java, C++, and the shell
• Model Catalog: browse models with pricing and context windows
• Rate Limits: limits, budgets, and how to handle them
• API Reference: full OpenAPI specification