New releases of modeling-app v6 and chap-core v2

Announcements - Annonces
,
1

Modelling App 6.2 Release Notes

Overview

Version 6.2 introduces a new prediction workflow built around reusable Prediction Setups. Instead of configuring every prediction from scratch, you can save a prediction configuration once, rerun it whenever needed, schedule it, and import results back into DHIS2.

The release also improves evaluation exploration, model selection, period selection, job monitoring, outbreak alert workflows, and in-app documentation.

Features available through the Modelling App

Reusable prediction workflows

Prediction configurations can now be saved as reusable Prediction Setups.

A setup stores your selected model, organisation units, period type, data mappings, and prediction configuration, allowing you to:

  • Rerun predictions with a single click.
  • Monitor previous prediction runs.
  • Schedule recurring predictions.
  • Reuse or edit import mappings.
  • Import prediction results directly into DHIS2.

This replaces the previous one-off prediction workflow and makes routine forecasting significantly easier.

Better evaluation analysis

Evaluation results are now easier to explore and compare.

New improvements include:

  • Richer filtering for custom plots.
  • Faster evaluation comparison pages.
  • Improved chart navigation with zoom and pan controls.
  • Unified downloads for datasets and evaluation metrics.

Improved model configuration

Creating evaluations is now simpler with:

  • Searchable model selection.
  • DHIS2-native period pickers.
  • Improved model information and readiness indicators.

Improved operations and support

The application now provides:

  • Better job monitoring and recent activity.
  • Prediction scheduling management.
  • Improved bug reporting with diagnostic information.
  • Expanded user guides throughout the application.

Quality improvements

  • Fixed stale analytics responses and uncertainty chart data.
  • Improved handling of archived models and long-running evaluation requests.
  • Various usability and performance improvements.
  • Expanded automated test coverage.

Compatibility

  • Supports DHIS2 2.41+.
  • Requires CHAP Core 2.0+.

CHAP Core 2.0–2.1 Release Notes

These releases provide the backend capabilities required by Modelling App 6.2 while also introducing new functionality for CLI users, model developers, and system administrators.

Features available through the Modelling App

  • New Prediction Setup architecture powering reusable prediction workflows.
  • Improved evaluation plotting and visualisations.
  • Faster API responses and backend performance improvements.
  • Support for weekly seasonal thresholds.

New CLI features

  • Causal analysis – Generate counterfactual predictions to explore “what if?” scenarios, estimating how disease outcomes might have changed under different hypothetical conditions such as increased intervention coverage or altered environmental factors.
  • New commands for evaluation aggregation, model schema inspection, and model card generation.
  • MLflow tracking – Optionally log evaluation runs to MLflow, making it easier to track experiments, compare model performance across runs, and maintain a reproducible history of model development.
  • Improved self-diagnostics through chap test.
  • Additional reporting and export capabilities.

Platform improvements

  • Automatic Chapkit service registration and synchronization.
  • Pluggable threshold framework.
  • Improved health checks and deployment support.
  • Enhanced OpenAPI documentation.
  • Faster CLI startup and gzip-compressed API responses.

Breaking changes

  • The deprecated evaluate and evaluate-hpo CLI commands have been removed.

For complete technical details, see:

Nouvelles versions de Chap 2.0/2.1