SDK Gears [NEW]

This page shows examples of writing a gear using the SDK.

A pre-requisite for using this portion of the SDK is reading the Flywheel gear spec, and understanding how the configuration is structured, especially with inputs and configuration values.

The flywheel.gear_context.GearContext class provides a simplified interface for performing common tasks in the lifecycle of a gear, such as accessing input files, configuration values, logging messages, accessing the SDK client and writing to the output folder.


The first step of using this interface is to create an instance of the GearContext class and, if desired, initialize logging. You can also print your configuration to the log to make troubleshooting easier during development:

import flywheel
with flywheel.GearContext() as context:
   # Setup basic logging

   # Log the configuration for this job

Accessing Config

Once initialized, the job configuration is accessible as a regular python dict from the context:

# Get the configured speed, with a default of 2
my_speed = context.config.get('speed', 2)

Accessing Inputs

You can get the full path to a named input file, or open the file directly:

# Get the path to the input file named 'dicom'
dicom_path = context.get_input_path('dicom')

# Open the dicom file for reading
with context.open_input('dicom', 'rb') as dicom_file:
   dicom_data =

If you have context inputs, you can get their values directly:

# Get the FSL license context value
fsl_license = context.get_context_value('fsl_license')
if fsl_license is None:
   raise RuntimeError('No license found!')

Accessing the SDK Client

If your gear is an SDK gear (that is, it has an api-key input), you can easily access an instance of the Flywheel SDK Client:

# Lookup a project using the client
project = context.client.lookup('my_group/Project 1')

Downloading BIDS

If your project or session is fully curated for BIDS, you can download all or a portion of the project or session to a working directory in BIDS layout.

# Download all files from the session in BIDS format
# bids_path will point to the BIDS folder
bids_path = context.download_session_bids()

# Download anat and func files from the project in BIDS format
# bids_path will point to the BIDS folder
bids_path = context.download_project_bids(folders=['anat', 'func'])

Writing Outputs

The path to the output directory is available as a variable on the context, and helper methods exist for opening an output file for writing:

print('Output path: {}'.format(context.output_dir))

# Open an output file for writing
with context.open_output('out-file.dcm', 'wb') as f:

Writing Metadata

Occasionally it’s useful to add or update metadata on the destination, one of the parent containers, or files on the destination (including output files)

This can be done using the metadata helper functions. The metadata will be written either when write_metadata() is called, or the context is exited (if using a with statement)

# Metadata will be written at exit of the "with" block,
# unless an exception occurs
with flywheel.GearContext() as context:
   # Update the session label
   context.update_container_metadata('session', label='Session 1')

   # Update the destination (e.g. acquisition) label and timestamp
   updates = {
      'label': 'fMRI_Ret_bars',
      'timestamp': '2014-05-07T08:50:07+00:00'

   # Set the modality and classification of an output file
   context.update_file_metadata('out-file.dcm', modality='MR', classification={
      'Intent': ['Functional'],
      'Measurement': ['T2*']