Dependency management on Anyscale
This page provides an overview of dependency management on Anyscale.
What features does Anyscale provide for dependency management?
Anyscale provides a suite of features to help you manage dependencies throughout your project lifecycle. Broadly, features related to dependency management fall into the following buckets:
- Stable container images provided by Anyscale.
- Building and using custom container images.
- Features for iterative development in Anyscale workspaces.
- Options to customize Anyscale jobs and services.
The following table provides an overview of Anyscale features related to dependency management:
| Feature | Description |
|---|---|
| Base images | Official Ray distributions packaged with common Python and system dependencies, maintained by Anyscale. See Anyscale base images. |
| Custom images | User-defined container images built on Anyscale or other third party tools. Typically extend Anyscale base images with custom packages, environmental variables, code, init scripts, and other custom logic. See Custom images on Anyscale. |
| Workspace containerfiles | Use containerfile syntax to iterate and build custom container images while working with Anyscale workspaces. See Iterate on workspace container images. |
| uv | Use uv to manage Python packages and projects during dev or production with workspaces, jobs, or services. See Use uv to manage Python dependencies. |
| Runtime environments | Use runtime environments to add custom dependency management at any level of your Ray application, such as for tasks and actors. See Runtime environments on Anyscale. |
These features provide you with flexibility to use the tools and patterns you're comfortable with for development. See Develop Anyscale applications.
For most use cases, Anyscale recommends packaging all code and dependencies as custom images before deploying workloads to production. See Container-driven development on Anyscale.