intura_ai Python SDK
Run the models you own on your own CPU
The intura-ai package downloads a model you own and runs it in your own process. After the download there is no network call, no per-request price, and your text never leaves your machine.
Your CPU
A model loads from a local cache in well under a second and predicts in your process: on a server, a laptop or an edge device.
No per-request cost
You buy a model once and get every version of it. Run it as often as you like.
Text stays with you
Only the download talks to Intura. What you predict on never leaves your machine.
From nothing to a prediction in three steps
- 1
Install the package
Python 3.10 to 3.13. It brings everything a prediction needs: onnxruntime (CPU), tokenizers, numpy and httpx.
Read the quickstartbashpip install intura-ai - 2
Create a download key
Get a model from the catalogue, then create a download key on your account page. The SDK reads it from INTURA_API_KEY.
Create a download keybashexport INTURA_API_KEY=sk_... - 3
Load a model and predict
The first load downloads the model into a local cache. Every load after that works from disk, offline and without a key.
What predict returnspythonfrom intura_ai import Intura model = Intura().models.load("chat-commerce-intent-id") model.predict("kak, paket saya kok belum sampai ya?") # {'label': 'shipping', 'score': 0.996}
Every page
Get started
Use a model
- Load and predictLoad a model with client.models.load, then predict and predict_batch. What each model's decision rule returns, and why batches run one text at a time.
- Versions and the cacheWhich version load picks, how to pin one or fetch the newest, where the cache lives, and how to download a model into a folder of your own.
- Your libraryList everything your account owns with client.library.list(), and the models you may download with client.models.list().
- AsyncAsyncIntura for an asyncio program: the download is awaited, the prediction is the same CPU-bound LocalModel.
Reference
- ErrorsEvery exception intura_ai raises, when each one happens, and what to do about it. A loaded model's predict raises none of them.
- ConfigurationThe client's arguments and environment variables: download key, cache folder, API root, timeouts, retries, your own httpx client and logging.
- Upgrading from 2.xintura-ai 3.0 only downloads the models you own and runs them. What replaced each 2.x call, type, error and CLI command.
Need a model to run? Each model page lists its slug, use cases and versions. Models