# Composable ML

> Composable ML - Learn how to use custom tasks and composable ML in DataRobot.

This Markdown file sits beside the HTML page at the same path (with a `.md` suffix). It summarizes the topic and lists links for tools and LLM context.

Companion generated at `2026-05-06T18:17:09.599566+00:00` (UTC).

## Primary page

- [Composable ML](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html): Full documentation for this topic (HTML).

## Sections on this page

- [Manage custom tasks](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html#manage-custom-tasks): In-page section heading.
- [Manage custom task versions](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html#manage-custom-task-versions): In-page section heading.
- [Create custom task version](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html#create-custom-task-version): In-page section heading.
- [List custom task versions](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html#list-custom-task-versions): In-page section heading.
- [Retrieve custom task version](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html#retrieve-custom-task-version): In-page section heading.
- [Update custom task version](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html#update-custom-task-version): In-page section heading.
- [Download custom task version](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html#download-custom-task-version): In-page section heading.
- [Prepare a custom task version for use](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html#prepare-a-custom-task-version-for-use): In-page section heading.

## Related documentation

- [Developer documentation](https://docs.datarobot.com/en/docs/api/index.html): Linked from this page.
- [Developer learning](https://docs.datarobot.com/en/docs/api/dev-learning/index.html): Linked from this page.
- [Python API client user guide](https://docs.datarobot.com/en/docs/api/dev-learning/python/index.html): Linked from this page.
- [Modeling](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/index.html): Linked from this page.
- [Specialized workflows](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/index.html): Linked from this page.
- [Manage execution environments](https://docs.datarobot.com/en/docs/api/dev-learning/python/mlops/custom_model.html#custom-models-execution-environments): Linked from this page.
- [CustomTaskFileItem](https://docs.datarobot.com/en/docs/api/reference/sdk/blueprints.html#datarobot.models.custom_task_version.CustomTaskFileItem): Linked from this page.

## Documentation content

Composable ML consists of two major components: [the DataRobot Blueprint Workshop](https://blueprint-workshop.datarobot.com/) and custom tasks, detailed below.

Custom tasks provide users the ability to train models with arbitrary code in an environment defined by the user.

For details on using environments, see: [Manage execution environments](https://docs.datarobot.com/en/docs/api/dev-learning/python/mlops/custom_model.html#custom-models-execution-environments).

## Manage custom tasks

Before you can upload code for a custom task, you need to create the entity that holds all the metadata.

```
import datarobot as dr
from datarobot.enums import CUSTOM_TASK_TARGET_TYPE

transform = dr.CustomTask.create(
    name="a convenient display name",  # required
    target_type=CUSTOM_TASK_TARGET_TYPE.TRANSFORM,  # required
    language="python",
    description="a longer description of the task"
)

binary = dr.CustomTask.create(
    name="this or that",
    target_type=CUSTOM_TASK_TARGET_TYPE.BINARY,
)
```

A task, by itself is an empty metadata container.
Before using your tasks, you need create a [CustomTaskVersion](https://docs.datarobot.com/en/docs/api/dev-learning/python/modeling/spec/custom_task.html#custom-task-versions) associated with it.
A task that is ready for use will have a `latest_version` field populated with this task.

```
binary.latest_version
>>> None

execution_environment = dr.ExecutionEnvironment.create(
    name="Python3 PyTorch Environment",
    description="This environment contains Python3 pytorch library.",
)
custom_task_folder = "datarobot-user-tasks/task_templates/python3_pytorch"
task_version = dr.CustomTaskVersion.create_clean(
    custom_task_id=binary.id,
    base_environment_id=execution_environment.id,
    folder_path=custom_task_folder,
)

binary.refresh()  # In order to see the change, you need to GET it from DataRobot
binary.latest_version
>>> CustomTaskVersion('v1.0')
```

If you create a new version, that will be returned as the `latest_version`.
You can download the latest version as a zip file.

```
binary.latest_version
>>> CustomTaskVersion('v1.0')

custom_task_folder = "/home/my-user-name/tasks/my-updated-task/"
task_version = dr.CustomTaskVersion.create_clean(
    custom_task_id=binary.id,
    base_environment_id=execution_environment.id,
    folder_path=custom_task_folder,
)

binary.refresh()
binary.latest_version
>>> CustomTaskVersion('v2.0')

binary.download_latest_version("/home/my-user-name/downloads/my-task-files.zip")
```

You can `get`, `list`, `copy`, exactly as you would expect.`copy` makes a complete copy of the task: new copies of the metadata, new copies of the versions, new copies of uploaded files for the new versions.

```
all_tasks = CustomTask.list()
assert {el.id for el in all_tasks} == {binary.id, transform.id}

new_binary = CustomTask.copy(binary.id)
assert new_binary.latest_version.id != binary.latest_version.id

original_binary = CustomTask.get(binary.id)

assert len(CustomTask.list()) == 3
```

You can `update` the metadata of a task.
When you do this, the object is also updated to the latest data.

```
assert binary.description == new_binary.description
binary.update(description="totally new description")

assert binary.description != new_binary.description
assert original_binary.description != binary.description  # hasn't refreshed from the server yet

original_binary.refresh()
assert original_binary.description == binary.description
```

And finally, you can `delete` only if the task is not in use by any of the following:

- Trained models
- Deployments
- Blueprints in the AI catalog

Once you have deleted the objects that use the task, you will be able to delete the task itself.

## Manage custom task versions

Code for Custom Tasks can be uploaded by creating a Custom Task Version.
When creating a Custom Task Version, the version must be associated with a base execution environment.
If the base environment supports additional task dependencies (R or Python environments) and the Custom Task Version contains a valid requirements.txt file, the task version will run in an environment based on the base environment with the additional dependencies installed.

### Create custom task version

Upload actual custom task content by creating a clean Custom Task Version:

```
import os

from datarobot.enums import CustomTaskOutboundNetworkPolicy

custom_task_id = binary.id
custom_task_folder = "datarobot-user-tasks/task_templates/python3_pytorch"

# add files from the folder to the custom task
task_version = dr.CustomTaskVersion.create_clean(
    custom_task_id=custom_task_id,
    base_environment_id=execution_environment.id,
    folder_path=custom_task_folder,
    outbound_network_policy=CustomTaskOutboundNetworkPolicy.PUBLIC,
)
```

To create a new Custom Task Version from a previous one, with just some files added or removed, do the following:

```
import os

import datarobot as dr

new_files_folder = "datarobot-user-tasks/task_templates/my_files_to_add_to_pytorch_task"

file_to_delete = task_version.items[0].id

task_version_2 = dr.CustomTaskVersion.create_from_previous(
    custom_task_id=custom_task_id,
    base_environment_id=execution_environment.id,
    folder_path=new_files_folder,
)
```

Please refer to [CustomTaskFileItem](https://docs.datarobot.com/en/docs/api/reference/sdk/blueprints.html#datarobot.models.custom_task_version.CustomTaskFileItem) for description of custom task file properties.

### List custom task versions

Use the following command to list Custom Task Versions available to the user:

```
import datarobot as dr

dr.CustomTaskVersion.list(custom_task_id)

>>> [CustomTaskVersion('v2.0'), CustomTaskVersion('v1.0')]
```

### Retrieve custom task version

To retrieve a specific Custom Task Version, run:

```
import datarobot as dr

dr.CustomTaskVersion.get(custom_task_id, custom_task_version_id='5ebe96b84024035cc6a6560b')

>>> CustomTaskVersion('v2.0')
```

### Update custom task version

To update Custom Task Version description execute the following:

```
import datarobot as dr

custom_task_version = dr.CustomTaskVersion.get(
    custom_task_id,
    custom_task_version_id='5ebe96b84024035cc6a6560b',
)

custom_task_version.update(description='new description')

custom_task_version.description
>>> 'new description'
```

### Download custom task version

Download content of the Custom Task Version as a ZIP archive:

```
import datarobot as dr

path_to_download = '/home/user/Documents/myTask.zip'

custom_task_version = dr.CustomTaskVersion.get(
    custom_task_id,
    custom_task_version_id='5ebe96b84024035cc6a6560b',
)

custom_task_version.download(path_to_download)
```

## Prepare a custom task version for use

If your custom task version has dependencies, a dependency build must be completed before the task can be used.
The dependency build installs your task’s dependencies into the base environment associated with the task version.

see: [Prepare a custom model version for use](https://docs.datarobot.com/en/docs/api/dev-learning/python/mlops/custom_model.html#custom-models-dependencies)