Sequential document loading during init() was the main bottleneck: each cached
doc required a serial IDB read + key derivation + decryption + Automerge.load.
With N documents this meant N× single-doc latency instead of 1×.
Changes:
- Add loadMany() to EncryptedDocStore for parallel Promise.all() batch loading
- Cache derived space/doc CryptoKeys in DocCrypto (one derivation per session)
- Add preloadSyncStates() to DocSyncManager for parallel sync state loading
- Update all 11 module local-first-client init() methods to use batch APIs
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add Automerge schemas, lifecycle hooks (onInit, docSchemas), and
local-first client wrappers for all remaining PG modules:
rWork, rVote, rCal, rFiles, rCart, rBooks, rTrips, rInbox, rSplat, rFunds.
Each module now:
- Defines typed Automerge document schemas (schemas.ts)
- Registers docSchemas and onInit hook with SyncServer reference
- Moves initDB() from top-level to onInit for unified startup
- Has a client-side local-first wrapper (local-first-client.ts)
Dual-write route handlers will be wired incrementally per module
following the rNotes pattern established in Phase 2.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Jeff Emmett