Organizations

Everything you upload belongs to an organization: your company, or a client's team you work with. One account can belong to several organizations; the switcher in the top bar moves between them, and the 11-digit number in the app's address says which one you're in. Each organization has its own files, people, model settings and credits, and nothing crosses from one to another.

Projects

A project holds the files of one site, block or flight. Its page lists the files with their state, the storage they use and the jobs running on them; its map shows them together. How many projects and how much storage you can use depends on your plan (Settings › Plans & Billing shows both meters).

What you can upload

File	Formats	What happens
Orthomosaics, multispectral and thermal imagery	`.tif`, `.tiff` (GeoTIFF, with its coordinate system)	Checked and turned into a cloud-optimized GeoTIFF, then shown on the map.
Inspection results, boundaries, assets	`.geojson`, `.json`, or a zipped shapefile (`.zip`)	Checked on upload; Process makes them ready for questions.
Photos and plain images	`.png`, `.jpg`, `.jpeg`	Stored with the project. Without coordinates they aren't placed on the map.

A shapefile goes in one ZIP with its .shp, .dbf, .shx and .prj files. The .prj matters: without a coordinate system we can't tell where the shapes are.

Large uploads

A single file can be up to 50 GB.
Files go from your browser straight to storage in parts, several at a time. If the page reloads or the connection drops, choose the same file again: the parts already stored are skipped.
An upload left unfinished for three days is cancelled and its reserved space freed.
When storage would go over your plan's limit, the upload is refused before it starts, with what's left.

What processing does to a vector file

Processing turns a GeoJSON or shapefile into a dataset the agent can query. It runs in the background with a live log, and you can leave the page meanwhile.

Every feature is read and moved to WGS 84 (EPSG:4326). Broken geometries are repaired, and the log says how many.
Each field gets a type from its values, so a column of numbers stored as text, with the odd "N/A", still counts as numbers.
Your organization's fast model looks at the fields and picks which one holds the class (like the defect type), the status and the id. If no model is available, built-in rules make the choice.
Areas and lengths are measured on the ellipsoid, so square metres stay right far from the equator.
The field values are indexed for search, on our own servers: nothing is sent out for that step.

Processing uses AI, so it needs a model source: Platform Models with credits, or your own key. The file's page shows what was found: feature count, geometry types, every field with its type and role, and the counts per class.

Archives we refuse

Before a ZIP is opened we check it: more than 2,000 files, more than 20 GB once unpacked, an implausible compression ratio, paths that point outside the archive, or encryption make us refuse it with the reason. When your server has virus scanning turned on, archives are also scanned, and a flagged one is never used.

Deleting files and projects

Before you delete a file, we show what goes with it: its search index, and anything still pointing at it, like a share link, a task or an automation. Those have to be revoked or removed first. Deleting a project removes its files from storage too.