Built around the PM interviews: one property at a time, mostly PDFs, scans, Excel, JSON,
and odd legacy files. Pick the property, add a short note if needed, and upload whatever files you got.
1. Intake context
This is stored locally in your browser for the intake UI. The backend just receives the selected property metadata for now.
2. Files
Drop files, folder contents, PDFs, Excel, zips, whatever
This is an internal intake surface. No file type restrictions. Use it for the messy reality the PM described.
Current flow: uploads are stored in R2, queued with property metadata, and processed through the same extraction pipeline. Use a small subset first if you want to inspect quality before uploading the full archive.