SpectrumX SDK Changelog

0.1.20 - 2026-06-02

• Fixes:
  • Fixed crash when owner and related fields are missing from API responses: the Capture.owner field (and User.name/User.email for cascading partial data) are now optional with None defaults. Previously, the SDK would raise a Pydantic ValidationError when the server omitted these fields from capture payloads during file and capture listing, causing the entire operation to fail.

0.1.19 - 2026-05-15

• Features:

  • Added start_time and end_time parameters to list_files and download: this gives users the ability to filter file directory downloads associated with DigitalRF captures based on a time span within the capture bounds (similar to time filtering in the web UI on SDS)
  • Added capture_uuids, top_level_dirs, and artifacts_only parameters to download_dataset: this allows users to input specific capture UUIDs or file directories associated with the dataset to download. artifacts_only allows user to filter for directly connected files on a dataset specifically. In this case, capture_uuids entry is ignored, but users may define top_level_dir file directories to filter artifacts for download on a large dataset. Users may run list_dataset_captures to see the UUID and top_level_dir of the captures on the dataset they wish to download.
  • Added list_dataset_captures method: this allows users to list all captures associated with a dataset.
  • Added list_dataset_artifact_files method: this allows users to list all artifact files associated with a dataset.

• Observability:

  • Added captures and files attributes to Dataset model: This allows visibility from the dataset side into attached captures and files.
  • Added capture_start_iso_utc, capture_end_iso_utc, capture_start_display, and capture_end_display fields to the Capture model: Users can now inspect the indexed capture time bounds from OpenSearch on capture metadata, both in UTC and in a display-formatted time zone.
  • Added gateway property to Client: Advanced users and test code can now access the underlying GatewayClient directly via sds.gateway.

• Housekeeping:

  • Paginator now surfaces API warnings: When the server returns warning messages in paginated responses, they are now logged automatically via the SDK's logger.

0.1.18 - 2026-04-30

• Features:

  • Added delete and revoke_share_permissions methods for datasets: this allows users to (soft) delete datasets in the SDS through the SDK and revoke ALL share permissions from datasets if needed before deletion or in general.
  • Added revoke_share_permissions and detach_from_datasets methods to captures: this gives users the ability to revoke share permissions or detach captures from connected datasets when they need to delete a capture.
  • Added detach_from_datasets methods to files: this gives users the ability to detach files from connected datasets when they need to delete them. Note: Files CANNOT be detached from captures. Delete a the parent capture FIRST to delete the file.

• Observability:

  • Added fields displaying ownership, share permission, and asset connection information to SDK models: this allows users to see more relevant information when retrieving or listing assets like whether they are shared, who they are shared with, who owns them, and what other assets the target is attached to (like files to captures and datasets).
    • New file attributes: owner, captures, datasets
    • New capture attributes: owner, is_shared, is_shared_with_me, share_permissions
    • New dataset attributes: owner, is_shared, share_permissions, datasets
  • Added new models for User and UserSharePermission: This allows for visibility into the users and share permissions connected to assets.

• Fixes:

  • Fixed upload failure when using relative local paths: the file discovery step in uploads broke when a relative path was given as local_path (e.g. sds.upload(local_path="my_dir", ...)). The root cause was passing the resolved (absolute) root path but feeding candidates from an unresolved (relative) rglob iterator, causing Path.relative_to() to fail with a ValueError. This was a regression introduced in v0.1.15 when switching to anyio.Path for async path operations.

0.1.17 - 2025-12-20

• Fixes:
  • Improved file downloads: more reliability for file downloads when they need to be resumed.
    • File downloads now use a temporary file during download to avoid partial files being left behind if the download is interrupted.
    • When overwrite is False and a local file would be overwritten, we skip re-downloading it.
    • When overwrite is True and the checksums don't match with server, we re-download and replace the local file to match the server's.
    • If local file is identical to server's, we now skip the download entirely, even when overwrite is True.

0.1.16 - 2025-12-16

• Fixes:
  • Added persist_state to upload_multichannel_drf_capture: users can now control whether to persist the upload state of files during multi-channel DRF captures.
• Reliability:
  • Relaxed validation in capture listing: capture listing is now more tolerant of cases where a given capture fails validation, now listing the ones that can be loaded correctly instead of failing the whole operations. Container fields like files now also default to empty ones if missing, instead of failing validation.

0.1.15 - 2025-12-02

• Features:
  • Added support for Python 3.14: the SDK is now compatible with Python 3.14. We're thus dropping support for version 3.10 in this and future releases.
  • Improved UX for file uploads: the new file upload process:
    • Backwards compatible with previous method.
    • Discovery step lists all files before starting upload.
    • Concurrent uploads with asyncio: up to 5 streams by default.
    • Progress tracking the number of files and bytes, in human-friendly units.
    • Tracking completed :white_check_mark:, failed :x: , in progress :hourglass_flowing_sand:️ and skipped:rabbit2: files. On Windows, we use ASCII characters for better compatibility.
  • New file upload persistence: we now keep a local copy of the upload state of files.
    • This makes resuming interrupted uploads faster. The persistence mode can be bypassed with persist_state=False in upload methods in case users want to force the file checks (previous behavior).
• Reliability:
  • Increased the battery of tests for file uploads and progress tracking.
  • Several new tests covering features of upload state persistence.

0.1.14 - 2025-10-07

• Documentation:
  • Added and SDK walkthrough notebook to the users' guide: SpectrumX SDK walkthrough.
• Fixes:
  • Reviewed logic to ignore files: now files like .DS_Store, *.tmp, and other patterns in .sds-ignore are correctly skipped for uploads.

0.1.13 - 2025-08-26

• Features:
  • Added download_dataset method to client: Users can now download datasets through the SDK by passing in a dataset UUID e.g. sds_client.download_dataset(uuid=<dataset_uuid>, to_local_path=...).
  • Added a files_to_download argument to the download() method: Can now supply a pre-filtered list of files to download rather than downloading all files in a path.

0.1.12 - 2025-08-12

• Features:
  • Added name parameter to upload_capture method: upload_capture() now accepts an optional name parameter to set custom capture names.

0.1.11 - 2025-07-03

• Features:
  • Added upload_multichannel_drf_capture SDK method to Client: Multi-channel captures can now be uploaded to the SDS.

0.1.10 - 2025-06-05

• Features:
  • Added created_at to captures:: captures now feature a created_at datetime attribute when this information is available.
  • Allowing upload of sigmf-data files: removed restriction of octet-stream files in uploads.

0.1.9 - 2025-05-08

• Features:
  • Capture deletions: captures may now be deleted using the SDK with sds_client.captures.delete().
  • Advanced searches: introducing preliminary support for advanced capture searches using OpenSearch queries. See the OpenSearch Query Tips for details about the syntax. The SDK endpoint is sds_client.captures.advanced_search(): the method's docstring has information on arguments and links to documentation.
• Usage improvements:
  • Allowing positional args: SDK methods that receive a single UUID as an argument now allow it to be passed as a positional argument e.g. sds_client.get_file(<uuid>) is equivalent to sds_client.get_file(uuid=<uuid>).
  • Capture uploads: new high-level method that combines the upload of a directory of files with a capture creation: sds_client.upload_capture().
  • Usage guide updates: the SDK usage guide now includes information on how to check the results returned by a directory upload.
  • Logging control:: client's verbose flag propagates to submodules and sub-APIs.
• Bugfixes and reliability improvements:
  • Bugfix: File.local_path attribute of some downloaded files was not correctly set.
  • Path usage tests:: new tests to cover regressions of different ways to use paths from SDK methods.
  • Capture read test: tests for client.capture.read() now check the number of files associated to the capture is correct.

0.1.8 - 2025-03-24

• Features and improvements:
  • #72 Added Windows compatibility.
  • #79 SDS files may now be deleted using the SDK.
  • #78 The capture_type argument of sds_client.captures.listing() is now optional.
  • #78 The method sds_client.captures.read() is now available for reading a single capture, complementing the existing sds_client.captures.listing() method.
  • #78 When retrieving a single capture with capture = read(<uuid>), the list of files associated to it is available in capture.files.
    • Note these are CaptureFile instances, which hold a subset of the information in an SDS' File instance. To get all metadata on a file, use sds_client.get_file(<uuid>).
• Fixes:
  • #77 Fixed incorrect file name being set when the file contents are already in SDS.
  • #77 local_path is now set for uploaded files across all three upload modes.
  • #71 File permissions are now included in file uploads; redesigned handling of permission flags.
• Housekeeping:
  • #76 Fixed documentation typos; Updated and expanded note on concurrent access in the usage guide.

0.1.7 - 2025-03-07

• Features:
  • spectrumx.api.captures.create() now accepts a scan_group for RadioHound captures.
  • index_name from spectrumx.api.captures.create() is dropped: the index is automatically chosen from the capture type.

0.1.6 - 2025-02-24

• New features:
  • Support for Python 3.10 and 3.11.

0.1.5 - 2025-02-10

• Breaking changes:
  • spectrumx.models.File moved to spectrumx.models.files.File.
• New features:
  • SDS Capture creation and listing.
• Housekeeping:
  • Refactored client by branching its method implementations and client configuration to separate modules.

0.1.4 - 2025-01-17

• New features:
  • File listing with lazy-loaded paginated results.
  • Introduced new file upload modes to avoid re-uploading files already in SDS.
• Housekeeping:
  • Usage documentation updated.
  • See tests/e2e_examples/check_build_acceptance.py for more SDK usage examples.
  • Refactoring of internal modules.
  • Improved test coverage.