AI Assistant
f8c19bb5b4
Root cause: even with SQL-level stripping, PostgreSQL must TOAST-decompress entire multi-MB JSONB values from disk before any processing. For 5 entries with PDF attachments (25-50MB total), this takes several seconds. Fix: store base64 attachment data in separate namespace 'registratura-blobs'. Main entries are inherently small (~1-2KB). List queries never touch heavy data. Changes: - registry-service.ts: extractBlobs/mergeBlobs split base64 on save/load, migrateEntryBlobs() one-time migration for existing entries - use-registry.ts: dual namespace (registratura + registratura-blobs), migration runs on first mount - registratura-module.tsx: removed useContacts/useTags hooks that triggered 2 unnecessary API fetches on page load (write-only ops use direct storage) Before: 3 API calls on mount, one reading 25-50MB from PostgreSQL After: 1 API call on mount, reading ~5-10KB total