Book Ingest — Phase 03: MDX → Supabase

Purpose

Land the validated corpus in Supabase so the e-reader, commerce flow, and pgvector RAG can serve from a single source of truth.

Schema (Drizzle, src/lib/database/schema/books.ts):

books — one row per book per language; embeds manifest (full book.json snapshot), commerce fields, reader defaults, RAG config.
book_chapters — one row per chapter; mdx_content text, content_sha256 for drift detection, is_preview for paywalled chapters, generated fts tsvector.
book_sections — display anchors (stable IDs the e-reader bookmarks against). Re-embedding does NOT regenerate sections.
book_chunks — RAG retrieval units (regenerable). HNSW pgvector index on embedding (cosine, m=16, ef_construction=64).
user_book_purchases | progress | bookmarks | highlights — user-scoped tables, RLS-protected.

$ARGUMENTS: