Changelog
Selected release notes are listed below. Current imports, defaults and limitations are documented in the API reference.
v0.7.2 (2026-10-19)
- Reuse movement caches by a 16-character data/settings fingerprint instead of the latest dated file. Changed requests preserve earlier results; old date-only caches are ignored without migration or rewriting.
- Forward inactivity-window and prior settings through
load_or_track_movement(); include the effective seed, normalized schedule and package version in cache keys. - Recompute only the matching cache with
use_cache=False, using atomic saves that preserve previous output on failure. Unreadable keyed files raise an error. - Preserve text identifiers, numeric precision and Monthly period-column spelling through cache round trips. Keep unkeyed artifact-helper naming compatible.
v0.7.1 (2026-10-18)
- Correct driver raw-event windows to matching calendar quarters, including the Q4 year rollover. Newly calculated driver counts can differ from v0.7.0.
- Respect
count_colon count axes in separate-model fitting; returned row counts and eligibility thresholds are unchanged. - Accept saved
last_active_quartermetadata in the original trajectory helper without a merge collision or caller-data mutation. - Provide an explicit release-tag annotation while preserving configured signing and the existing push confirmation.
- Use net raw-period changes for driver contributions, add decreases and coverage diagnostics, and report undefined percentages for zero net change. Validate duplicate movement rows, explicit columns and numeric bins.
- Preserve saved results; no model refit or migration is performed by viewing.
Documentation follow-up
- Separate installation from the README and add a ready-to-open IT incidents AutoViz example using existing saved results, with a workspace screenshot.
- Reconcile methodology, score/event rules, API examples, return values, file formats, cache reuse and saved-result column guidance with current code.
- Remove unsupported accuracy, lead-time and threshold-calibration claims; the subsequent v0.7.1 changes above correct the documented driver limitations.
v0.7.0 (2026-10-18)
- Add an axis-free tile-and-point logo beside the intact PriorityX name and retain its embedded tab icon in self-contained HTML exports.
- Add validated, portable saved-result bundles and explicit CSV import with schema, duplicate-field and numeric-value validation.
- Add a loopback-only saved-results workspace with packaged Plotly.js charts: linked quadrant matrix, YTD transition timeline and recent trajectories. Quarterly opens when available; type labels remain caller-defined.
- Fit browser timeline labels automatically and provide equal chart panels, exact data tables, linked selection, zoom/reset and portable view settings.
- Add self-contained interactive HTML and bounded Matplotlib PNG/SVG exports over shared saved-result/view contracts. Display controls do not fit models or change saved analytical values; existing Matplotlib defaults remain available.
- Preserve caller-provided entity display labels without domain inference.
- Require Python 3.12 or later; support for Python 3.10 and 3.11 is removed.
- Default source-checkout development to Python 3.12 and configure CI for Python 3.12, 3.13 and 3.14 on Linux, macOS and Windows.
- Check installed-wheel app launch, spawned exports and disconnected HTML reopening in CI; require the quality matrix before release builds.
- Start the numeric loopback workspace without reverse-DNS resolution, avoiding launch stalls when the host resolver is unavailable.
- Refresh and include the lockfile when preparing a version bump so locked installation checks remain valid before tagging.
- Record the installed package version as the producer of newly created bundles; retain the recorded producer when reopening older results.
- Keep static trajectory endpoint labels clear of visible paths and axes, and include remaining line crossings in layout reports.
- Document exact saved-result columns, identity/date mapping, leading-zero IDs, local launch on macOS/Windows and the analysis-to-AutoViz data boundary.
v0.6.0 (2026-01)
Simplified Visualization Path API
- Added
plot_pathandcsv_pathparameters toplot_priority_matrix(),plot_transition_timeline(),plot_entity_trajectories() - Path implies save: providing path automatically saves output
- Accepts Path objects directly (no
str()wrapper needed) - Old API (
save_plot=True+plot_dir+plot_filename) deprecated
v0.5.0 (2026-01)
Simplified API Returns
- Added
return_stats,return_metadata,return_pathparameters to core functions - Functions now return just DataFrame by default; pass
return_*=Truefor tuple - Renamed
track_cumulative_movement()→track_movement()(old name deprecated) - Fixed threshold calibration docs and x_score/y_score interpretation
v0.4.0 (2025-01)
Unified API with Axis Metrics
- Introduced
x_metric/y_metricparameters tofit_priority_matrix()for flexible axis definitions beyond default volume/growth - Added
x_score/y_scorecoordinate-based quadrant classification (replaces Random_Intercept/Random_Slope naming in outputs) - Added Gamma family support for continuous GLMM (e.g., duration or amount metrics)
- Gaussian GLMM support for metric-based axes
- Improved quadrant label rendering with configurable legend position (
legend_loc,show_legend) - Row-aligned quadrant label offsetting to avoid legend overlap
Trajectory Auto-Selection
- Added
highlight_top_nparameter toplot_entity_trajectories()for automatic entity selection - Added
highlight_byparameter:"trajectory_distance"(euclidean path) or"total_movement"(manhattan) - Added
recent_periodsparameter to limit trajectory window to last N periods
Enhanced Driver Analysis
- Added
numeric_colsparameter toextract_transition_drivers()for amount/duration breakdowns- Explicit bins:
{"amount": [0, 1e6, 5e6, 10e6]} - Auto quantiles:
{"duration": 4}(4 bins)
- Explicit bins:
- Renamed
quarter_from/quarter_totoperiod_from/period_to(backwards compatible) - Renamed
top_n_subcategoriestotop_n,min_subcategory_deltatomin_delta(backwards compatible) - Added
save_txt,txt_path,txt_modeparameters todisplay_transition_drivers()for file output
v0.3.0 (2024-12)
Entity Metrics and Priority Indices
- Implemented
add_priority_indices()for composite index calculation:- RI (Risk Index): volume + growth + severity
- SQI (Service Quality Index): resolution speed + recovery rate
- EWI (Early Warning Index): blends RI and SQI for prioritization
- Added
aggregate_entity_metrics()helper for per-entity metric aggregation - Added
sensitivity_analysis()for weight stability testing - Monthly granularity support (beta) in GLMM and visualizations
- Unified
px.*namespace imports pattern
v0.2.0 (2024-11)
Transition Tracking and Visualization
- Implemented
track_movement()for entity trajectory tracking (originallytrack_cumulative_movement, renamed in v0.5.0) - Added
extract_transitions()for quadrant movement detection - Transition timeline visualization with priority-based coloring (Crisis/Investigate/Monitor/Low)
- Spike markers (
*X,*Y,*XY) for volume/growth spikes - Deterministic seeding via
PRIORITYX_GLMM_SEEDfor reproducible GLMM fits plot_entity_trajectories()for movement path visualization- Transition driver analysis (experimental)
v0.1.0 (2024-10)
Initial Release
- Core GLMM estimation with Poisson mixed effects (statsmodels backend)
fit_priority_matrix()main API for entity prioritization- Quadrant classification (Q1-Q4) based on volume and growth
- Priority matrix scatter plot visualization
- Quarterly, yearly, and semiannual temporal granularity
- Date filtering and minimum observation thresholds
- CSV and plot export with configurable output directories