Issue 19:
- entity_id now optional on POST /api/import/schedule
- Resolution: explicit entity_id > name match from row 1 > auto-create
- New params: create_entity_type (default fund), create_vintage_year
- Dry-run reports resolution (will_create/matched/existing)
- Frontend: radio toggle between 'Create from file' and 'Use existing'
Issue 20:
- GET /api/entities/rollup: invested_cents and last_signed_value_cents
per entity in one pass (SQL aggregates, no waterfall)
- EntitiesList uses rollup instead of N+1 API calls
- Parse Carta XLSX natively (openpyxl) instead of CSV-only
- Column map confirmed against Fund I export: Investment/Asset/
Investment date/Shares/Cost/Value/Last Valuation date
- Handle grouped layout: company subtotal rows + indented position rows
- SAFEs: shares=0 in source → null in DB (not additive)
- Money already in dollars in XLSX cells, convert to cents directly
- Dates parsed from Excel datetime objects
- CSV fallback preserved for other formats
- Re-import guard catches both seed and non-seed existing rounds
- Tested: 14 holdings, 17 positions, total value $8,527,204.58 matches
- Dockerfile: multi-stage build (Node frontend + Python backend)
- StartOS manifest, interfaces, backups, version graph
- start.sh: auto-generates session secret, creates first approver on boot
- Backend serves built frontend as SPA in production
- Single data volume for SQLite DB, platform backups cover it
Johnny 5