Skip to content

Changelog

Notable changes to BigClust, newest first. Versions follow semantic versioning; the format is loosely based on Keep a Changelog.

For the complete commit-level history, see the releases on GitHub.

Unreleased

Fixed

  • Meta refresh counts are now reported as values (row/column intersections) alongside the number of rows they sit in — "cells" read as the neurons themselves. The share that merely populated a newly added column is called out separately, since that fills every matched row at once and can dwarf the real edits.
  • A value whose source representation differs but which converts back to what is already stored (a source sending "111" for an integer column) is no longer counted as a change. It never reached the table, but it was reported on every single run.

Changed

  • Widgets are opened per embedding. In a project with several embeddings, View → Connectivity Table / Distance Heatmap / Feature Comparison are now submenus listing the embeddings, so a widget no longer has to be reached by first switching to the embedding that happens to hold its data. Embeddings without the required data are listed but disabled, saying what they lack; the entry itself is greyed out only when no embedding has it. Projects with a single embedding are unchanged, and the keyboard shortcuts still open the widget for the active embedding. See the View menu.
  • import bigclust2 no longer loads the GUI stack. Qt, pygfx and wgpu are imported the first time the app itself is started, so the Python helpers (ProjectBuilder, update_project_meta) and bigclust2 update now work — and start noticeably faster — on machines with no display or GPU.

Added

  • Headless meta-data refresh — bigclust2 update /path/to/project pulls current annotations from the project's declared sources and writes them into the project, no window required. --auto-map matches columns by name, --dry-run shows what would change. Also available as bigclust2.update_project_meta() from Python. See the command line reference and refreshing meta data.

0.2.0 — 2026-07-25

A large feature release centred on new interactive widgets, a from-scratch project-building workflow, and this documentation site.

Added

  • Project builder (File → Build Project) — a guided dialog for turning raw data into a BigClust project without hand-writing the info file. See the how-to guides.
  • Find Nearest widget (Selection → Find Nearest...)— search for a neuron's nearest neighbours in feature space, and grow the current selection by nearest neighbour. See Widgets.
  • Command palette (press Cmd+Shift+P) to search and run any command from the keyboard. See also Keyboard and mouse.
  • Custom metadata — bring your own metadata sources and merge them into a loaded dataset. See Refresh the meta data.
  • Credential handling — centralized management of service credentials for the annotation backends (Clio, neuPrint, FlyTable). See Credentials.
  • Local skeleton sources in 3D viewer
  • Update checker — notifies you when a newer release is available.
  • Center on selection — new menu item and keyboard shortcut (Shift+C).
  • Documentation site — getting-started guides, how-to guides, concepts, and this reference section.

Changed

  • Smarter label placement, with better collision handling and positioning.
  • Reworked and substantially expanded connectivity widget.
  • Improved point visuals in the scatter view.
  • Filtering (the scoping system) can now match empty and null values.

Fixed

  • The search widget no longer crashes on a trailing comma.

0.1.1 — 2026-07-17

Added

  • Debug menu: new Traceback option.

Changed

  • Raised the minimum nglscenes version.

0.1.0 — 2026-07-16

Initial public release on PyPI.

Core capabilities at first release: interactive exploration of clusterings of high-dimensional connectomic data, embedding and feature comparison, the connectivity widget, growing and shrinking selections, pushing annotations to backends, working across multiple views, and Neuroglancer export.