Breaking Changes
- The `input_data` property of `Invocation` now returns an `Optional[InvocationInput]`. It used to return a `list` of file name patterns.
Features Added
- Added `extra_headers_provider` param to `HttpClient` which accepts a callback which generates headers on every HTTP request at runtime. This was added so we ourselves can send internal trace IDs.
- Added getters for `created`, `created_at`, `modified`, `modified_at` to `Dataset`, `File` and other related entities, eliminating the need to access them via `.record`.
- Enhanced `Dataset::download_files` to return both `FileRecord` objects and their corresponding local save paths.
- Added `ActionRuntime::get_input` method to inspect resolved input data references during Action execution.
- Introduced `InvocationInput` - a richer way of specifying inputs to action invocations which will not be limited to dataset file paths. Full platform support will be delivered in stages.
- Introduced `Layouts` - a way to create a saved arrangement of panels for visualizing data in Roboto.
- Added `TriggerRecord::causes`, `CreateTriggerRequest::causes`, and `UpdateTriggerRequest::causes` to allow triggers to filter which evaluation causes they respond to.
- Added `TriggerEvaluationCause::FileIngest` to allow triggers to respond to when a file is marked as `IngestionStatus::Ingested`.
Bugs Fixed
- Added dynamic import guard around `roboto.version` in `requester.py` to fix in-IDE tests from Roboto's development environment (before `version.py` is dynamically generated)
- Don't reject API requests that contain extra fields. This enables backwards-compatibility with outdated SDK builds and forwards-compatibility for adding new fields to the SDK independent of our server release cycle.