使用 Homebrew 安装 dstack

项目链接

• 主页

命令和别名

• dstack

项目历史

The package trail starts on PyPI in September 2021, while the GitHub repository was created in January 2022. The project evolved from a reproducible ML workflow tool into a broader GPU orchestration system covering development environments, training tasks, services, fleets, volumes, and backends.

By 2026 the README described dstack as a unified control plane for GPU provisioning and orchestration across GPU clouds, Kubernetes, and on-prem clusters, with support for NVIDIA, AMD, Google TPU, and Tenstorrent accelerators. The docs and README also show the project adapting to agentic workflows through dstack skills for tools such as Claude, Codex, and Cursor.

采用历史

Distribution is source-backed through PyPI and Homebrew. PyPI metadata records hundreds of releases and version 0.20.26 in June 2026, while the Homebrew formula uses the same release from the Python source distribution.

Homebrew analytics are modest compared with general-purpose infrastructure tools, but the formula matters because it gives macOS and Linux users a package-manager path into a Python ML orchestration stack without starting from a source checkout.

使用方式

Typical usage starts with installing the CLI or server, running `dstack server`, adding a project with a server URL and token, and applying YAML configurations from a repository. The official README documents `.dstack.yml` workload files and user or server configuration under `~/.dstack`.

The package is aimed at teams that want one workflow for development environments, batch tasks, training jobs, services, and GPU fleet management across different compute backends.

为什么软件包爱好者会关心

dstack is notable as part of the modern wave of ML infrastructure CLIs that package cloud, Kubernetes, GPU scheduling, and developer workflow into a single local command. It sits in the same package-manager lane as tools that developers install before a full platform is available.

For package nerds, its interesting wrinkle is the mix of Python packaging, Homebrew distribution, server configuration, repo-local YAML definitions, and rapidly changing AI accelerator support.

时间线

• 2021-09: PyPI records the first dstack release artifacts.
• 2022-01: The current GitHub repository is created.
• 2025-12: dstack 0.20.0 release line emphasizes fleet-first UX and events.
• 2026-06: PyPI and Homebrew package dstack 0.20.26.

Related projects

• Kubernetes is one of the supported orchestration backends.
• PyPI is the upstream package distribution channel used by the Homebrew formula.
• Claude, Codex, and Cursor are named in the README as agent clients that can use dstack skills.

来源

• https://github.com/dstackai/dstack
• https://dstack.ai/docs
• https://pypi.org/pypi/dstack/json
• https://formulae.brew.sh/api/formula/dstack.json