Search

Search components, tokens, patterns, architecture

Patterns

Explainability

“Why?” is answered in place. The explanation layer (Layer 3) lives inside each card's drawer — capacity drivers, signals used, assumptions — so the user never leaves her context to interrogate the system.

Rules

  1. 01Every engine output exposes primary and secondary drivers; the drawer renders them in priority order.
  2. 02Explanations are sentences, not feature lists.
  3. 03The explanation names the trade-off when modes conflict (symptoms override cycle optimism).

Examples

What Changed drawer

“Capacity ↓ 12 — mostly short sleep (5.1h), plus yesterday's HIIT. Cycle context unchanged.”

Anti-patterns

What breaks this pattern

  • A global “How Peri works” page as the only explanation surface.
  • Showing driver weights as raw numbers.

Do

  • ·Explain at the moment of surprise
  • ·Order drivers by actual contribution

Don't

  • ·Make users hunt for reasoning
  • ·Expose internal scores without interpretation