Cobalt 0.3.9 is available now. Update via: pip install --upgrade cobalt-ai
Key Features
The release introduces statistical flexibility in group comparisons. Beyond the standard t-test, users now have access to "the Kolmogorov-Smirnov test and the Wilcoxon rank-sum test" plus permutation-based p-value approximations.
A new Workspace.get_group_neighbors() method enables discovery of neighboring data subsets, which can serve as comparison groups in the UI.
Graph Improvements
The layout algorithm received substantial enhancements:
- Node overlap prevention for clearer relationship visualization
- Components organized by size from left to right
- Automatic rotation for optimal space utilization
- Threaded animation preventing UI interference
Examples visualizing before and after v0.3.9:
Quality Updates
Embeddings for text data can now be created directly by calling CobaltDataset.add_text_column_embedding(). The licensing prompt was clarified, and GroupResultsCollection now inherits from GroupCollection, streamlining failure group analysis.
Bug Fixes
A timestamp display issue affecting graph coloring was resolved.