# Customize Python runtime environment


## Build Custom Runtime Images

Zerops allows you to build custom runtime images (CRI) when the default base runtime images don't meet your Python application's requirements. This is an optional phase in the [build and deploy pipeline](/features/pipeline#runtime-prepare-phase-optional).

## Configuration

### Default Python Runtime Environment

The default Python runtime environment contains:

- <span>Alpine 3.20</span>
- Selected version of Python when the runtime service was created
- [zCLI](/references/cli)
- Git and PIP

### When You Need a Custom Runtime Image

If your Python application needs more than what's included in the default environment, you'll need to build a custom runtime image. Common scenarios include:

:::important
You should not include your application code in the custom runtime image, as your built/packaged code is deployed automatically into fresh containers.
:::

Here are Python-specific examples of configuring custom runtime images in your `zerops.yml`:

### Basic Python Setup

### Using Build Files in Runtime Preparation

For complete configuration details, see the [runtime prepare phase configuration guide](/features/pipeline#configuration).

## Process and Caching

### How Runtime Prepare Works
The runtime prepare process follows the same steps for all runtimes. See [how runtime prepare works](/features/pipeline#how-it-works) for the complete process details.

### Caching Behavior
Zerops caches custom runtime images to optimize deployment times. Learn about [custom runtime image caching](/features/pipeline#custom-runtime-image-caching) including when images are cached and reused.

### Build Management
For information about managing builds and deployments, see [managing builds and deployments](/features/pipeline#manage-builds-and-deployments).

:::warning
Shared storage mounts are not available during the runtime prepare phase.
:::

## Troubleshooting

If your `prepareCommands` fail, check the [prepare runtime log](/python/how-to/logs#prepare-runtime-log) for specific error messages.