Install dvc with Homebrew, chocolatey, Nix, scoop, winget

Project links

• Homepage
• Repository
• Upstream docs
• Package manager page

Commands and aliases

• dvc

Project history

The public GitHub repository was created in March 2017, and the oldest release metadata exposed by GitHub shows a beta release, 0.8.1, in May 2017. The project developed around the idea that code, data, model files, metrics, and pipeline stages should be reproducible from a repository without forcing large binary artifacts into Git itself.

DVC 2.0.0 was released on 2021-03-03, reflecting a period when the tool had broadened from data-file tracking into experiment and pipeline workflows. The 3.x release line was active through 2026, with release metadata showing frequent 3.x updates and a 3.67.1 release on 2026-03-31.

Adoption history

DVC became common in machine-learning engineering because it mapped data-science needs onto a familiar developer mental model: Git commits track code and lightweight metadata, while DVC remotes hold larger artifacts. Package-manager coverage across Homebrew, Chocolatey, Scoop, winget, Nix, and Python packaging reflects that it is used by both local workstation users and CI-oriented teams.

Its adoption also followed the rise of reproducible ML pipelines and experiment tracking. Instead of being only a storage helper, DVC became a CLI layer for project structure: `dvc init`, `dvc add`, `dvc push`, `dvc pull`, `dvc repro`, and experiment commands give data-science repos repeatable operational verbs.

How it is used

Typical users initialize DVC inside an existing Git project, add datasets or model files to DVC tracking, configure a remote object store, and commit the generated metadata files to Git. Pipelines are then described so stages can be reproduced and cached, while results and metrics can be compared across experiments.

DVC is especially useful when teams need a Git-like review and checkout workflow for data without storing gigabytes in Git history. It is also used in CI to fetch exact data/model versions and run reproducible training or evaluation jobs.

Why package nerds care

DVC is package-nerd significant because it is one of the clearest examples of a developer-tool CLI crossing into data science without abandoning Unix and Git habits. It treats datasets as content-addressed artifacts, keeps metadata text-friendly, and lets package managers install the same `dvc` command across macOS, Linux, and Windows.

It also sits in a crowded but important lineage that includes Git LFS, lakeFS, MLflow, Pachyderm, and experiment trackers. DVC's distinction is that it keeps a repo-centered workflow rather than forcing every team into a hosted platform first.

Timeline

• 2017-03-04: Public GitHub repository created.
• 2017-05-04: Oldest visible GitHub release metadata shows beta release 0.8.1.
• 2021-03-03: DVC 2.0.0 released.
• 2023-2026: DVC 3.x became the long-running active release line.
• 2026-03-31: GitHub release metadata shows DVC 3.67.1.

Related projects

• DVC is related to Git and Git LFS for version-control workflows, object stores such as S3-compatible remotes for artifact storage, and ML platforms such as MLflow and lakeFS that overlap with reproducibility, experiment, or data-versioning concerns.

Sources

• https://doc.dvc.org
• https://github.com/treeverse/dvc
• https://api.github.com/repos/treeverse/dvc
• https://api.github.com/repos/treeverse/dvc/releases