- Added Snapshot merging / manipulation
- Class for highlighting extractions onto source PDFs and adding table of contents.
- Row Association now also sorting on 'bbtop'
- PDF manipulation features
- Support for classification predictions
- Dependency installation
- Find from questionnaire ID added to finder class.
- ModelGroupPredict support added.
- Added module to get metrics for all models in a group.
- Multi color highlighting and annotations added for PDF highlighting.
- Added stagger dataset upload feature for large doc datasets.
- Added default retry functionality for certain API calls.
- Added additional snapshot features.
- Wait kwarg added to submit review method.
- Better support for dataset creation / adding files to teach tasks.
- Classes for model comparison and model improvement.
- Plotting functionality for model comparison.
- Positioning class added to assist in relative prediction location validation
- Added # of samples labeled to metrics class.
- Teach classes in indico_wrapper
- New line plot for number of samples in metrics class.
- Update to highlighting class with new flexibility and bookmarks replacing table of contents.
- Abillity to include metadata with highlighter
- Ability to split large snapshots into smaller files
- Updated functionality for large dataset creation. Batch options allow for more reliable dataset uploads.
- New distance measurements in the prediction Positioning class.
- New Features on the Extractions class: predictions that are removed by any method are saved in an attribute if they're needed for logs, etc.; get all text values for a particular label; get most common text value for a particular label.
- Better exception handling for Workflow submissions and more flexibility on format of what is returned (allows custom response jsons to avoid the WorkflowResult class).
- Updated metrics plot to order ascending based on latest model
- New feature in Positioning class to calculate overlap between two bounding boxes on the same page
- Optional dependencies to support M1 installation
- New feature in FileProcessing class to read and return file as json
- New feature in Highlighter class to redact and replace highlights with spoofed data
- New Download class to support downloading resources from an Indico Cluster
- Upgrades client to 5.1.3 and upgrades SDK calls for Indico 5.x compatibility
- FindRelated class in indico_wrapper
- Upgrades client to 5.1.4
- New feature to now support staggered looped learning
- Ground truth compare feature to compare a snapshot against model predictions and receive analytics
- Modifies IndioWrapper class to updated CreateModelGroup call to work with Indico 5.x
- Updates Snapshot class to account for updated target spans
- Updates Add Model calls to aligh with 5.1.4 components
This is the first major version release tested to work on Indico 6.X.
- Refactored AutoReview to simplify setup.
- Replaced AutoClassifier with AutoPopulator to make ondoc classification model training simple. This class also includes a "copy_teach_task" method that is a frequently needed standalone method.
- Simplified a StaggeredLoop implementation to inject labeled samples into a dev workflow (deprecated previous version).
- Added support for unbundling metrics.
- Added the
Strucure
class to support building out workflows, datasets, teach tasks. As well as to support copying workflows.
- Small but important fix to add original filename to the workflow result object
- Removed staggered loop support and removed highlighting support.