Changelog

[v0.2.1] – 2025-01-21

Status panel worker display now shows "16 workers" before processing starts instead of misleading "0/16 workers"
Help panel README handling now uses importlib.metadata to load main README from package metadata in pip-installed environments

Streaming mode with StreamingOrchestrator for in-memory cutout processing using shared memory, enabling direct processing without disk I/O
Flux-conserved resizing using the drizzle algorithm to preserve photometric accuracy during image resampling
Parquet input support allowing source catalogues to be provided in Parquet format in addition to CSV
Raw cutout extraction (cutout_only mode) for outputting unprocessed cutouts directly from FITS tiles
External FITSBolt configuration support with TOML serialization for seamless integration with FITSBolt pipelines
Log level selector dropdown in the UI header for runtime log verbosity control
Vulture dead code detection CI workflow to identify and prevent unused code accumulation
Ruff import sorting CI check to enforce consistent import ordering across the codebase
Comprehensive benchmarking suite (paper_scripts/) for performance evaluation and reproducibility of paper results
Async streaming example (examples/async_streaming.py) demonstrating programmatic streaming mode usage

Catalogue streaming architecture with CatalogueStreamer enabling memory-efficient processing of catalogues with 10M+ sources through atomic tile batching
Default output folder changed from cutana/output to cutana_output for cleaner project structure
Default resizing mode changed to symmetric for more intuitive cutout dimensions
Logging configuration now follows loguru best practices: disabled by default, users opt-in via logger.enable("cutana")
WCS handling optimized for FITS output with correct pixel scale and WCS - no SIP distortions implemented
Documentation updated for Euclid DR1 compatibility with improved README and markdown formatting
Source mapping output now written as Parquet instead of CSV for better performance with large catalogues
Dependencies updated: fitsbolt>=0.1.6, images-to-zarr>=0.3.5, added drizzle>=2.0.1, scikit-image>=0.21

WCS pixel offset corrected 1-based indexing and half-pixel offset issues affecting cutout positioning
Subprocess logging resolved ANSI escape codes and duplicate log folder creation
Windows compatibility fixed temp file permission issues in streaming mode tests
Parquet file selection in UI now properly filters and displays Parquet files
Flux conservation integration properly applied in cutout_process_utils.py
Normalisation bypass allowing "none" config value to skip normalisation entirely

10x memory reduction for large catalogue processing through true streaming implementation
WCS computation optimisation reducing overhead for FITS output generation
Single-threaded FITSBolt mode for improved stability in multi-process environments

JobCreator class and associated dead code identified through vulture static analysis
Obsolete example notebooks (Cutana_IDR1_Setup.ipynb, backend_demo.ipynb) replaced with updated documentation