Pillar — Knowledge

Summary

Curated, stage-aware knowledge cards + AI Q&A on top. Every card has an indian-context-toggle.

Canonical values

Key	Value	Last changed
PILLAR.ID	knowledge	2026-07-27
PILLAR.MVP.CARD-COUNT	~200	2026-07-27
PILLAR.CARDS.STRUCTURE	title, TL;DR, body, global-view, india-view, sources, related-cards	2026-07-27
PILLAR.CARDS.FORMATS	text (all cards) · audio narration (~80% of cards) · short video (~20% of cards, 2–5 min)	2026-07-27
PILLAR.AUDIO.NARRATORS	~4 curated narrators, warm human voices; Hindi + English tracks per card	2026-07-27
PILLAR.VIDEO.MAX-DURATION	5 min	2026-07-27
PILLAR.AI.ENTRY-POINT	Raised center tab-bar button ("Ask AI") on every screen	2026-07-27
PILLAR.AI.MODEL	Claude Sonnet 4.5 for chat, Haiku 4.5 for classification	2026-07-27
PILLAR.AI.CITATION-REQUIRED	Yes — chat answers cite Cards, not the model	2026-07-27
PILLAR.AI.REFUSAL	Hard refusal for no-medical-advice categories	2026-07-27
PILLAR.FREE.LIMIT	3 AI chats/week · read/listen unlimited (all cards)	2026-07-27
PILLAR.PAID.LIMIT	Unlimited AI · unlimited watch of video cards · offline audio	2026-07-27

Related

• content-taxonomy — Card structure and tag taxonomy
• no-medical-advice — safety rules on AI
• indian-context-toggle — every Card has this toggle
• surface-mobile-knowledge-feed, surface-mobile-knowledge-card-detail, surface-mobile-knowledge-ai-chat

Sources

• Design decision