Get started
Quickstart
Install intura-ai, create a download key, then load a model you own and run your first prediction on your own CPU.
Install
pip install intura-ai
# or
uv add intura-aiThe package supports Python 3.10 to 3.13. It installs everything a prediction needs: onnxruntime (CPU), tokenizers, numpy, and httpx for the download. The old local extra still resolves, so pip install "intura-ai[local]" keeps working, but it adds nothing.
Get a model and a download key
- Get a model from the model catalogue. Buying a model once gets you every version of it.
- Create a download key on your account page. The key is shown once, so copy it then.
- Give it to the SDK in
INTURA_API_KEY:
export INTURA_API_KEY=sk_...See Download keys for the other ways to pass it, and for when you don't need one.
Load a model, then predict
from intura_ai import Intura
client = Intura() # reads INTURA_API_KEY
model = client.models.load("chat-commerce-intent-id")
model.predict("kak, paket saya kok belum sampai ya?")
# {'label': 'shipping', 'score': 0.996}The first load downloads the model into a local cache. After that it loads from disk in well under a second, with no network call and no key. Load once per process and keep the model.
A model that finds personal data returns spans instead of a label:
model = client.models.load("guardrails-pii-detection-id")
model.predict("Nama saya Budi Santoso, NIK 3171234567890001")
# [{'type': 'PERSON', 'start': 10, 'end': 22, 'value': 'Budi Santoso'},
# {'type': 'NATIONAL_ID', 'start': 28, 'end': 44, 'value': '3171234567890001'}]What predict returns depends on the model's decision rule. Load and predict has all five.
Try it from a shell
The package also installs an intura-ai command:
intura-ai library # what your account owns
intura-ai predict chat-commerce-intent-id "kak, paket saya kok belum sampai ya?"The Command line page has every command and flag.
Next
- Find a model's slug: Your library
- Pin a version, or put a model in a folder of your own: Versions and the cache
- Use it in an asyncio program: Async