API Reference

Client

garden_ai.GardenClient 

Main class for interacting with the Garden service.

Provides public get_garden method, and handles authentication with Globus Auth.

garden_ai.GardenClient.get_garden 

get_garden(doi: str) -> Garden

Return the published Garden associated with the given DOI.
Parameters:
doi: The DOI of the garden. Raises an exception if not found.
Returns:
The published Garden object. Any Modal functions in the garden can be called like methods on this object.

Garden objects

garden_ai.gardens.Garden 

Represents a collection of related functions, providing a way to organize and invoke machine learning models.

This class is geared towards users wishing to access a published Garden, and is meant to be instantiated by the client's get_garden method.

Attributes:

Name	Type	Description
`metadata`	`GardenMetadata`	The Garden's published metadata, including information such as title, authors, description, and DOI.
`modal_functions`	`list[ModalFunction]`	The callable functions associated with this Garden. Individual functions are also accessible like attributes on this object.

Example

Functions can be accessed as attributes of the Garden instance, allowing for intuitive calling of the associated functions:

client = garden_ai.GardenClient()
garden = client.get_garden("my_garden_doi")
result = garden.my_function(data)

garden_ai.gardens.Garden.get_job_status 

get_job_status(job_id: str) -> JobStatus

Get status information for a submitted HPC job.

Parameters:

Name	Type	Description	Default
`job_id`	`str`	Globus Compute task ID returned by HpcFunction.submit()	required

Returns:

Type	Description
`JobStatus`	JobStatus object with current status and outputs

Example

status = garden.get_job_status(job_id)
print(status.status) # "completed"
if status.status == "completed":
... results = garden.get_results(job_id)

garden_ai.gardens.Garden.get_results 

get_results(job_id: str, output_path: str | Path | None = None) -> Any

Retrieve results from a completed HPC job.

Parameters:

Name	Type	Description	Default
`job_id`	`str`	Globus Compute task ID	required
`output_path`	`str \| Path \| None`	Optional local path to save results (for file-based results)	`None`

Returns:

Type	Description
`Any`	Job results (type depends on the function)

Raises:

Type	Description
`RuntimeError`	If job is not completed or failed

Example

results = garden.get_results(job_id)

Or save to file:

results = garden.get_results(job_id, output_path="./results.xyz")

garden_ai.gardens.GardenMetadata 

Represents the metadata defining a Garden.

Not meant to be instantiated directly by users. Instead, new Gardens should be created via the CLI or web UI

Attributes:

Name	Type	Description
`doi`	`str`	The Digital Object Identifier (DOI) for the Garden.
`title`	`str`	The title of the Garden.
`authors`	`UniqueList[str]`	A list of authors of the Garden.
`contributors`	`UniqueList[str]`	A list of contributors to the Garden. Defaults to an empty list.
`doi_is_draft`	`bool \| None`	Indicates if the DOI is in draft status. Defaults to True.
`description`	`str \| None`	A brief description of the Garden. Defaults to None.
`publisher`	`str`	The publisher of the Garden. Defaults to "Garden-AI".
`year`	`str`	The year of publication. Defaults to the current year.
`language`	`str`	The primary language of the Garden. Defaults to "en" (English).
`tags`	`UniqueList[str]`	A list of tags associated with the Garden. Defaults to an empty list.
`version`	`str`	The version of the Garden. Defaults to "0.0.1".
`owner_identity_id`	`UUID \| None`	The UUID of the Garden's owner. Defaults to None.
`id`	`int \| None`	An internal identifier for the Garden. Defaults to None.