Page Bundles (apps-pages)

AGILab can optionally launch page bundles (sometimes called “apps-pages”): standalone dashboards that complement the built-in workflow pages.

In the UI, page bundles are presented the same way as core pages, but they run in their own sidecar web process and are embedded back into the main app.

What is a page bundle?

A page bundle is a small web bundle project stored on disk and launched on demand in its own Python interpreter:

  • It lives under ${AGILAB_PAGES_ABS} (default: src/agilab/apps-pages).

  • It is discovered when it is a Python file directly under AGILAB_PAGES_ABS or a directory exposing src/<module>/<module>.py (or main.py / app.py).

  • It can ship its own .venv (or venv); otherwise Analysis will fall back to the shared locations referenced by AGILAB_VENVS_ABS and AGILAB_PAGES_VENVS_ABS.

For fastest setup, AGILAB Analysis also exposes a minimal template generator:

  • Choose a page name in the Analysis Create from template panel and click Create.

  • The generator creates <page>/pyproject.toml and <page>/src/<page>/<page>.py so the page is immediately discoverable.

  • The page title is derived from the page name.

  • You can also duplicate an existing page bundle as a starting point with Clone from existing apps-page.

  • Open the generated page and replace the sample logic with your own visuals.

Enabling bundles (per project)

Bundles are enabled per project by writing their module names into app_settings.toml:

[pages]
view_module = ["view_maps_network", "view_barycentric"]

You can edit this manually (PROJECT → APP-SETTINGS) or use Analysis → Configure, which writes the same list for you.

Included page bundles

This page lists the page bundles shipped with the repository. You can discover and launch them from ANALYSIS.

view_autoencoder_latentspace

Autoencoder-powered dimensionality reduction.

  • Input: exported dataframe (typically created in Orchestrate).

  • Output: latent-space plots, with colouring and train/test controls.

view_barycentric

Barycentric simplex visualisation for KPI-style features that sum to 1.

  • Input: dataframe with aggregated proportion columns.

  • Output: interactive simplex plot for relative contributions.

view_maps

2D map viewer for geolocated datasets.

  • Input: CSV/parquet with latitude/longitude columns.

  • Output: interactive map with sampling, palette, and basemap controls.

view_maps_3d

3D cartography view (Deck.gl) with optional overlays.

  • Input: one or more geolocated datasets.

  • Output: 3D map with extrusion/colour controls and layer toggles.

view_maps_network

Network topology viewer synchronised with geographic views.

  • Input: node positions + link definitions in the dataset.

  • Output: map + graph views to inspect connectivity, link types, and snapshots.