Python SDK

Use the Arize Python package to create datasets, run experiments, get traces, and log evaluations to Arize. You can also use our ML features to log predictions and actuals. See our API reference below.

Arize Python SDK Reference — Arize API Reference 7.38.1 documentation

Installing the package

pip install arize

Changelog

The Arize SDK has additional functionality that can be installed with extra dependencies:

client_python/CHANGELOG.md at main · Arize-ai/client_pythonGitHub

End of Support Table

Major Release	First Released	Latest	Support
7.x	June, 2023	latest	Ends June 1st, 2026
6.x	January, 2023	6.1.3	Ends June 1st, 2025
5.x	August, 2022	5.5.0	Ends November 1st, 2024
4.x	March, 2022	4.2.2	Ended June 1st, 2024
3.x	September, 2021	3.4.0	Ended April 1st, 2024
2.x	March, 2021	2.2.1	Ended July 1st, 2023
1.x	July, 2020	1.2.1	Ended March 1st, 2022
0.x	March, 2020	0.0.20	Ended March 1st, 2022

Available packages

The main package is arize, and you can add additional functionality by adding extra dependencies. is the minimum required for the additional packages.

Package	What It's For
arize	The primary Arize package for interfacing with Arize APIs.
arize[Tracing]	Components for logging traces and running experiments for LLM applications.
arize[AutoEmbeddings]	Automatically generate embeddings vectors for your predictions and actuals for ML models.
arize[NLP_Metrics]	Calculate evaluation metrics for your NLP Generative tasks.
arize[MimicExplainer]	Produce SHAP values using the surrogate explainability approach.