MIP combines the following independently released repositories. These component versions make up the MIP 9.2.1 release.
| Repository | Version | Role and major changes |
|---|---|---|
| platform-ui | 2.1.1 | Redesigned Experiment Studio and Data Handling, filter-based categorical columns, K-means preprocessing, refreshed visualizations, experiment folders, exports, and notebook entry point. Saving an experiment stores the reviewed run without executing it again. |
| platform-backend | 10.2.0 | Analysis and specification endpoints, experiment folder persistence and management, authenticated access for web and notebook analyses, kept runs that can be saved without re-execution, and one experiment API for preview, run and save. |
| exaflow | 1.2.1 | Federated execution and preprocessing, reusable K-means cluster columns, quartiles, binned Mann-Whitney U, Standardized Mean Difference, and updated analysis requests. |
| mip-jupyter | 0.2.1 | JupyterLab workspace, mip Python client (distributed as mip-client), welcome notebook, JupyterHub and single-user images, and Cohort Scout notebook support agent. Example notebooks now come from the shared, read-only mip-examples repository. |
- Deployment package: local and federated installation tooling in this repository.
- MIP Infrastructure: MIP KaaS infrastructure based on ArgoCD deployments.
- Notebook integration notes: JupyterHub, authentication, workspace images, and assistant configuration.
The backend 10.x and Exaflow 1.2.x releases change the analysis request/API contracts. Integrations must use the revised analysis and specification APIs; older clients and deployment configurations need review before upgrading. Notebook access requires JupyterHub deployment and authentication wiring, and Cohort Scout requires its assistant service to be configured.