1 slot open · booking from dec 2026→
messy spreadsheet → clean records

ImportEngine

2026

A data import engine: messy files in, validated records out to other systems.

For teams that receive spreadsheets they can't trust. Each file is profiled, its columns mapped and every row validated, then clean records are delivered to other systems.

ImportEngine pipeline: upload, profile, map, transform, validate, review and deliver

screens

Column mapping: messy headers such as 'tracking #' and 'Wt (kg)' mapped onto target fields, with auto-mapping suggestionsValidation results for a 248-row fileA 20,000-row run in the transform phase, processed by queued chunk jobs with live progress

key decisions

engine apart
The generic engine is fenced off from the host app: dependencies point one way, and architecture tests enforce the boundary.
big files
Large files fan out into chunked, queued jobs; the wizard shows live progress.
safe delivery
Delivery goes only to destinations that pass the same checks when saved, tested and used.
written rules
Architecture rules recorded with the reason behind each one, so agents follow them too.

where ai comes in

ai in the build
  • ›A QA skill where an agent uses the real UI as a scripted user and writes up what it found
  • ›Phased plans with review gates

stack

laravelreacttypescriptpostgresrediss3