v1.1.3

This is a minor release with many minor fixes and smaller improvements, mainly aimed at functionality and fixes for the january webinar series. Some of these changes are experimental.

Highlights

• New chap init command - Scaffold new models quickly with a single command
• UvRunner - Support for uv-managed Python environments
• ISO week format - Weekly time periods now use ISO-like format for chapkit models
• Preference learning CLI - New preference-learn command for model selection via A/B testing
• Evaluation improvements - New export-metrics, plot-backtest, and PDF report generation commands

Features

• Auto-start chapkit service for directory-based models
• Read supported_period_type from chapkit model info
• Add historical context to evaluation plots
• YAML plotting support with monthly/daily data in peak_diff metrics
• Xarray serialization for Evaluation class and evaluate2 CLI
• New example dataset for CLI testing

Bug Fixes

• Fix geojson not attached to dataset when fetching from database
• Sort prediction data by location and time_period
• Calculate deterministic metrics on median of samples (CLIM-295)
• Ignore runs directory when copying model directory
• Fix GitHub URL parsing with trailing slashes
• Pin pydantic-geojson to fix PointModel issues

Documentation

• Migrated from Sphinx to MkDocs
• Restructured CLI documentation
• Added preference learning and evaluation workflow guides
• Automated documentation testing with mktestdocs