Assets, Visualization & Dashboard Helpers

assets/ is a support layer

The assets package is not where recommendation models live. It contains reusable reporting/dashboard helpers and bundled presentation resources.

src/cooprecsys/assets/
├── icon/
├── dashboard_utils.py
├── statsrender.py
├── vizdata.py
├── vendors_css.zip
└── vendors_js.zip

dashboard_utils.py

Provides small transformations for human-readable labels, numeric formatting, gauge detection, scorecards, and chart normalization.

These helpers are useful when a model/report produces metrics such as NDCG, MAP, coverage, latency, fallback rate, or score distributions and the dashboard needs a stable JSON-friendly representation.

statsrender.py

Gen_MiniStats() converts tabular statistics into compact dashboard/report data. It belongs after model evaluation, not inside the training algorithm itself.

vizdata.py

This module contains data-level visualization helpers such as score distributions, 2-D embedding projections, similarity heatmaps, top-K similar-item calculations for diagnostics, and conversion of scores into prediction frames.

The similarity helper here is a visualization/analysis utility. It should not be confused with the Ary2Tower residual recommendation fallback, which deliberately uses a global Bayesian/recency prior rather than item-item similarity.

Packaged frontend resources

vendors_css.zip and vendors_js.zip are packaged frontend assets used by the broader dashboard/report environment. Keep them versioned with the code that expects their structure.

Example: prepare scorecard data

from cooprecsys.assets.dashboard_utils import generate_scorecards

cards = generate_scorecards({
    'ndcg@10': 0.61,
    'coverage': 0.84,
    'fallback_rate': 0.07,
})

The output can then be handed to the dashboard/report layer without coupling the UI to the internals of a model trainer.