v1.7.0

Summary

• Improved single-cell extraction scaling and memory control

  • Auto-calibrates buffered worker results against a RAM budget
  • Adds target_ram_utilization, max_inflight_result_batches, and flush_every
  • Streams multiprocessing results instead of materializing large lists
  • See: src/scportrait/pipeline/extraction.py

• More flexible segmentation configuration

  • Supports combining multiple nucleus or cytosol channels via max-intensity projection
  • More reliable overrides for segmentation channel IDs
  • Enables cytosolOnly on single-channel inputs
  • Safer serialized sharding plans and window handling
  • See:
    • docs/pages/workflow/segmentation_workflow.rst
    • src/scportrait/pipeline/segmentation/workflows/_base_segmentation_workflow.py

• Expanded SpatialData plotting

  • Coordinate-system aware image plotting
  • Direct plot_shapes support
  • Improved plot_labels for vectorized layers and annotation tables
  • Palette/group access enhancements
  • More robust segmentation-mask overlays
  • See:
    • src/scportrait/plotting/sdata.py
    • src/scportrait/plotting/init.py

• New image preprocessing helper

  • Added value_range_normalization
  • Improved normalization and downsampling logic
  • See: src/scportrait/processing/images/init.py

• New H5SC writer helper

  • Public numpy_to_h5sc(...) API for writing .h5sc files from NumPy arrays + metadata
  • See:
    • src/scportrait/io/h5sc.py
    • src/scportrait/io/init.py

• Stronger Phenix parsing

  • Improved handling of missing tiles with optional checks
  • More stable tile clustering under stage coordinate jitter
  • Deduplication for combined exports
  • See: src/scportrait/tools/parse/_parse_phenix.py

What's Changed

• [VERSION] Bump version to 1.6.2-dev0 (merge after release) by @github-actions[bot] in #361
• Revert "[VERSION] Bump version to 1.6.2-dev0 (merge after release)" by @sophiamaedler in #362
• Switched from tuple to list in config example by @nik-as in #366
• [VERSION] bumped spatialdata-plot version to 0.2.14 to fix incorrect channel scaling in default plotting by @sophiamaedler in #367
• Updated copyright year in docs by @nik-as in #370
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #371
• Revert "Revert "[VERSION] Bump version to 1.6.2-dev0 (merge after release)"" by @sophiamaedler in #373
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #374
• fix intermittent dataset download failures in CI/docs and improve robustness, typing, and documentation in the pipeline base classes by @sophiamaedler in #379
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #376
• improve documentation build process by removing warnings by @sophiamaedler in #380
• improve robustness of remote dataset retrieval by @sophiamaedler in #381
• improve image normalization capabilities by @sophiamaedler in #372
• Improve segmentation workflow by @sophiamaedler in #385
• [FIX] reduce peak memory usage during single-cell extraction by @sophiamaedler in #384
• [FIX] fix KeyError in Stitcher when do_intensity_rescale='full_image' by @sophiamaedler in #382
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in #383
• Add function to write h5sc files from AnnData Objects by @sophiamaedler in #387
• Phenix parser by @sophiamaedler in #386
• [VERSION] Bump version to 1.7.0 by @github-actions[bot] in #388

Full Changelog: v1.6.1...v1.7.0