Local saved-results workspace
The local workspace opens validated saved results without fitting models. It uses Plotly.js for linked quadrant matrices, transition timelines and trajectories. The existing Matplotlib plotting API and its defaults are unchanged.
The workspace requires Python 3.12 or later.
Install it in a virtual environment as shown in the installation instructions. Then run:
priorityx app --bundle results.json
# Or use the Python module:
python -m priorityx app --bundle results.json
Omit --bundle to open an empty workspace. --port 0 chooses a free port;
--port 8765 chooses a particular port. --no-browser prints the address without
opening it. The service binds to 127.0.0.1 and keeps one bundle in memory.
Keep this terminal open; press Ctrl+C to stop the service. For an unactivated
Windows environment, use .\.venv\Scripts\python.exe -m priorityx app.
It serves application assets only. It does not expose a filesystem browser,
write analytical outputs automatically, or provide remote hosting.
For a ready-to-open dataset, follow the IT incidents walkthrough. It uses saved example results and includes all three chart families.
Open and inspect
Use Open results → Open result bundle for a version 1 result bundle, or Open results → Import CSV files for explicitly assigned matrix, movement and transition roles. Import opens a dialog; closing it preserves the entered fields. Supply type, cadence, coverage and source metadata. CSV import validates required columns, duplicate headers, exact row widths and numeric values. Additional columns remain caller metadata; source-specific mapping happens in the caller’s adapter. A failed import leaves the current workspace intact.
Choose the data and columns
AutoViz reads saved analytical results, not a raw event CSV. The normal flow is
your event data → fit_priority_matrix() / track_movement() →
extract_transitions() → result tables or a bundle → AutoViz. Fitting and
movement tracking are explicit Python analysis operations; opening/importing
results and applying layout only regenerate visualizations.
During analysis, entity_col selects your source identifier/name column and
timestamp_col selects your source date column. In saved results, entity
provides both the display text and the identifier used to link charts. There is
no separate per-entity display-name selector. entity_type identifies the group;
type_label only changes its label in Type.
Movement and transition tables have their own period columns; the matrix gets
its snapshot date from metadata. See required columns
and the source-column mapping example.
The import dialog assigns file roles and metadata; it does not guess identity
or date columns. coverage_start describes the complete saved history, and
as_of the latest saved snapshot. Neither is a request to filter/refit data.
For a bundle created from your saved CSVs, follow the
CSV-to-bundle example, save it as
results.json, then launch python -m priorityx app --bundle results.json.
Alternatively, import the same three files using the dialog. Supplying only a
matrix enables its chart; add movement for trajectories and transitions for the
timeline. Keep the full saved movement table for event classification.
Inspect the linked charts
Type labels come from the bundle. Quarterly opens when available; a bundle with only Monthly results opens Monthly. The timeline shows January/Q1 through the latest saved period in the snapshot year, including blank periods. Preparation uses full movement history before display filtering. Missing tables or selected entities without fitted points/paths are identified beside the affected chart.
Selecting an entity links all charts. Type changes clear selection and reset the view; cadence changes retain the entity when it exists in the other cadence. Chart navigation and zoom do not make service requests or fit models. Exact matrix, event and trajectory values are available in View data, with Matrix, Timeline and Trajectories tabs. Each chart’s Data link opens its matching tab. Arrow keys, Home and End move between tabs. Timeline period headings remain visible during vertical scrolling.
Without an explicit saved trajectory selection, the workspace shows up to three entities with the largest total absolute trend changes over the displayed path window. Both rises and falls count. The default window is eight quarters or 12 months; equal scores use entity identifiers as a deterministic tie-breaker. An explicit saved list keeps its order. Details explains which selection mode is active. Selecting an entity shows its individual path; clearing selection returns to the default group. Named keys and endpoint letters identify the shown paths; line patterns distinguish paths that share a quadrant color. The trajectory table opens on the shown paths; Show → All saved entities includes all saved paths within the same period window. See the saved-result selection contract.
Layout and continuity
Layout opens a settings dialog and controls equal panel dimensions. Closing the dialog retains unapplied edits; a visible reminder keeps them discoverable. The timeline fits the panel width automatically: names wrap, period labels shorten when needed and crowded event groups use separate lanes within their entity row. Every YTD period and event remains available without horizontal timeline scrolling. Exact values remain in View data → Timeline below the charts. Apply layout before exporting changed settings. Measured placement wraps matrix labels and separates trajectory endpoint labels without changing saved coordinates. Unresolved conflicts remain visible and are included in the layout report.
Reset view restores chart ranges and timeline position. Reset settings restores the complete display defaults. Browser storage retains selection, actual axis ranges, scroll positions, data tabs, trajectory table scope and disclosures for each bundle. Menus and dialogs close on reload or export; older version 1 settings with separate table folds still open the corresponding data tab. Save settings and Load settings, inside Layout, provide an explicit portable round trip; settings for another bundle or unsupported schema are rejected.
Exports and offline operation
Export → Save offline HTML embeds the pinned Plotly basic 4.1.2 asset, result bundle, prepared views and settings. It reopens without a server or network, including type/cadence controls, linked selection, zoom, data tables and settings save/load. The file contains full saved history beyond the visible YTD window. Matrix SVG export also works offline. Treat exported data with the same care as the source bundle.
The Export menu also provides matrix SVG and print PNG/SVG. Print exports require the local Python service. They use the same chart populations, saved scores, event classes/suffixes and trajectory window, with a separate print profile; browser zoom does not crop the print population. Automatic sizing follows bounded, deterministic dimension candidates. Manual dimensions win. Minimum chart text is 12 pt. Remaining clipping/overlaps are reported, with every analytical mark retained. Exports have a 30-second render deadline, and concurrent print requests are rejected with a retry instruction. Downloads from an older view are discarded if the view changes during rendering.
The new opt-in renderer is also available directly:
from priorityx.results import ResultBundle
from priorityx.views import prepare_bundle, resolve_settings
from priorityx.viz.export import render_static
prepared = prepare_bundle(ResultBundle.load("results.json"))
settings = resolve_settings(None, prepared)
dataset = next(d for d in prepared["datasets"]
if d["entity_type"] == settings.entity_type
and d["cadence"] == settings.cadence)
fig, report = render_static(dataset, settings, "timeline")
fig.savefig("timeline.png", dpi=settings.dpi)
print(report.to_dict())
Build boundary and qualification
Application HTML/CSS/JavaScript, Plotly and its MIT license live inside the Python
package. priorityx.app.build_html() reads those assets directly; there is no Node runtime,
CDN fetch or dependency on development prototypes. Build distributions with
uv build; no additional Python runtime dependencies are required.
The CI matrix covers Linux, macOS and Windows on Python 3.12–3.14. After building, each job replaces the editable install with the wheel and checks console launch from a path containing spaces/non-ASCII text, both cadences, all chart families’ spawned PNG/SVG exports, and direct offline-file reopening in headless Chrome or Edge. Receipts and exported artifacts accompany each run. These automated checks do not certify physical touch, platform font appearance or Safari behavior. Future private hosting needs its own access and deployment design; the local launcher deliberately does not expose a remote bind option.