Files
geointel/docs/V1_SCOPE_FREEZE.md
T
Jens faeb58ef6d
GeoIntel release gates / Compile, test, contracts and builds (push) Successful in 1m49s
GeoIntel release gates / Python and npm vulnerability policy (push) Successful in 21s
GeoIntel release gates / Production AI image, SBOM and container scan (push) Successful in 5m39s
GeoIntel release gates / Deploy exact gated revision to Unraid (push) Failing after 58m43s
Initial public release
2026-08-31 21:56:53 +02:00

3.0 KiB

V1 Scope Freeze

V1 identity

GeoIntel V1 is the first usable GeoAI Workbench foundation. It must prove the full data-processing loop from project to geospatial output.

V1 must include

Project foundation

  • Create/list/open projects.
  • Add one or more areas.
  • Persist project metadata.
  • Persist area geometry.

Dataset manager

  • Upload raster datasets.
  • Upload vector datasets.
  • Extract metadata.
  • Persist dataset records.
  • Show import status.

Map workbench

  • Display selected area.
  • Display vector layers.
  • Display raster preview or placeholder tile/extent if full tiling is not ready.
  • Toggle layer visibility.
  • Adjust opacity.
  • Inspect feature properties.

Raster foundation

  • Read metadata using Rasterio/GDAL-compatible stack.
  • Clip raster by area if georeferenced.
  • Generate tiles or tile records for inference.
  • Store derived artifact paths.

Vector foundation

  • Import GeoJSON and zipped shapefiles.
  • Normalize CRS.
  • Validate/fix simple geometries.
  • Clip by project area.
  • Calculate area/length in metric CRS.

Reference data foundation

  • Support GRB/reference building layer through provider abstraction.
  • Support local fixture provider for offline demos.
  • Support OSM provider as fallback/context if external access is available.

Detection foundation

  • Implement detection run entity.
  • Implement inference provider interface.
  • Add a simple YOLO/Ultralytics provider or stub provider only if clearly marked as development provider.
  • Convert outputs to persisted detection geometries.
  • Export detections as GeoJSON.

Segmentation foundation

  • Implement segmentation run entity.
  • Implement segmentation provider interface.
  • Support mask artifact path.
  • Polygonize segmentation masks when georeferenced and feasible.

QA/QC foundation

  • Compare prediction geometries with reference geometries.
  • Calculate IoU-based matching.
  • Calculate precision, recall, F1.
  • Persist quality-check record.
  • Show matched/unmatched layers.

Exports

  • GeoJSON export for detections/segmentations/change outputs.
  • JSON metadata export.
  • Simple report artifact.

Tests

  • Backend unit tests for geometry metrics.
  • Backend unit tests for QA matching.
  • API smoke tests for projects/datasets/analysis runs.
  • At least one fixture-based workflow test for Demo 1 if possible.

V1 should include if time permits

  • Basic NDVI/NDWI/NDBI calculation for suitable rasters.
  • Simple vector change detection.
  • Simple dashboard cards for metrics.
  • Demo fixture loader.

V1 must not include

  • Full LiDAR workbench.
  • Model training studio.
  • Full MLOps registry.
  • QGIS plugin.
  • Multi-user authentication.
  • Complex PDF designer.
  • Real-time monitoring.
  • Production deployment hardening beyond local Docker.

V1 UX rule

Every page must have a useful empty state and a real connected state. No page should exist only as a design placeholder.

V1 quality rule

If a feature cannot be made geospatially correct yet, it must explicitly show a warning and avoid exporting misleading coordinates.