Data about a place, seen together.

Sentinel-1, Sentinel-2, Sentinel-3 and Sentinel-5P, through Copernicus Data Space Ecosystem. In code the collections are called exactly `sentinel-1-grd`, `sentinel-2-l2a`, `sentinel-3-slstr` and `sentinel-5p-l2`. We do not have our own satellites and we do not buy commercial high-resolution images — if you need 30 cm, that is a different discussion and a different provider.

No, and no one can give you that with Sentinel. Sentinel-2 passes over the same place roughly every 5 days, Sentinel-1 every 6–12 days, and the product needs processing time. That is why the default window of the layers is the last 7 days, ending yesterday. What you receive is the best available scene in the requested interval, with its declared date.

Optics is lost. We filter scenes with over 40% cloud cover and mask the pixels marked by the SCL band as cloud, shadow, cirrus or snow — they appear transparent, not invented. When optics cannot help, Sentinel-1 remains: radar passes through clouds and works at night, at the cost that its reading is different — it sees roughness and moisture, not colour.

No. That requires InSAR: pairs of scenes, precise orbits, DEM and phase unwrapping. In code this product is listed with the status `requires_pair_processing` — meaning declared as not implemented, not hidden. What we do instead: screening. There is a landslide triage module that presents itself as `flowmind-landslide-screening-v0`, heuristic, with the warning written in the API response that it is not confirmed detection.

Few, and it is better to hear it from us. A proprietary OSINT survey from July 2026 found a ceiling of approximately 34 live public road cameras across the whole country — 12 ANDSA plus 22 at the border. At today’s check, out of 116 surveyed streams, 13 were responding, all refreshed JPEG images, zero video streams. The rest are declared for the following reasons: expired subscription, rotated token, missing folder, inaccessible port. If someone promises you 150 cameras in Moldova, ask for the list.

No, and there is no way to do so in the current form. The 245 traffic lights are mapped as points in the data, from OpenStreetMap — the platform sees them on the map, it does not command them. Active control would mean integration with the physical controllers and safety responsibility, which is a separate project, with other partners.

No. There is no trained model in the repository and no machine learning code is called by the application; the `ml/gnn` and `ml/forecasting` directories are empty. The RL and GNN architecture in ARCHITECTURE.md is the target, marked as such in that very file.

The standard Copernicus account comes with 30,000 processing units and 30,000 requests per month, free of charge. Each requested tile uses some of them. The platform reduces consumption through cache (30 minutes in memory, 24 hours in response) and through limits on request size and interval, but it does not have a quota counter — if the installation has high public traffic, metering and a paid plan are things to decide beforehand, not afterwards.