aistore

What it is and what it does

AIStore is a Python SDK for interacting with AIStore clusters—a distributed object storage system designed for petascale AI and machine learning workloads. The package provides client-side APIs to create and manage buckets, upload and retrieve objects, and perform ETL operations on stored data. It includes native PyTorch integration for efficient data loading in training pipelines and a Botocore patch for S3-compatible access patterns.

The SDK depends on 13 common libraries (requests, pydantic, msgspec, tenacity, and others) to handle HTTP communication, data validation, serialization, and retry logic. It targets developers working with large-scale distributed storage, machine learning infrastructure, and high-performance computing environments where object storage efficiency matters.

Use it for:

• Load training data from AIStore clusters directly into PyTorch for distributed ML training.
• Manage and query petascale object storage buckets and objects programmatically from Python.
• Perform ETL operations on objects stored in AIStore clusters with built-in retry logic.
• Access AIStore via S3-compatible APIs using the included Botocore patch for existing workflows.
• Build data pipelines that fetch and process objects from distributed AIStore clusters.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides Python client APIs and PyTorch integration to access and manage AIStore clusters, buckets, and objects for distributed object storage operations.

Yes, if you are working with AIStore clusters or need distributed object storage integration in Python. The package is actively maintained, has no known vulnerabilities, installs with low friction, and is licensed permissively. Most valuable for machine learning and high-performance computing teams.

Install

aistore on PyPI

pip install aistore

uv add aistore

poetry add aistore

Installing aistore

Before you install

Low install friction with a pure-Python wheel and 13 well-established runtime dependencies. Active maintenance with recent commits and a 1912-star repository indicates ongoing development.

License in practice

MIT license permits unrestricted use, modification, and distribution in both open-source and commercial projects with minimal restrictions.

Quickstart

pip install aistore

from aistore.sdk import Client

client = Client()
buckets = client.list_buckets()

Requires a running AIStore cluster; Python 3.8 or later.

Verify before relying

• Whether PyTorch integration requires PyTorch as a separate install or is optional within aistore.
• Specific performance characteristics or throughput limits for typical workloads.
• Whether Botocore patch is automatically applied or requires manual activation.
• Default endpoint configuration and connection requirements for client initialization.

Package facts

Evidence: aistore-1.25.0-py3-none-any.whl

Keywords: AIStore, Artificial Intelligence, Deep Learning, ETL, High Performance, Lightweight Object Storage, Object Storage, Petascale

Tags

More Scientific/Engineering packages