Skip to main content
Beta — The Predict API is in beta. Request and response schemas may change between versions. Contact support@footstep.ai for access and feedback.

Predict API

The Predict API generates spatial probability distributions for missing person scenarios. Given a search area, subject profile, and environmental conditions, it returns per-hex probabilities across an H3 grid — telling you where to look first.

Models

Footstep Nova family

Footstep Nova is a family of spatial prediction models purpose-built for missing person scenarios. Nova models combine terrain analysis, land use classification, points of interest, weather conditions, and time-of-day signals to predict the most likely locations for a subject based on their behavioural profile. Nova supports multiple behavioural profiles. Each model accounts for how different subject types interact with their environment — a despondent individual moves differently to someone with dementia.

footstep-nova-1

The first release in the Nova family. Nova 1 delivers profile-aware, weather-aware spatial predictions across flexible search areas.

Key capabilities

  • Profile-aware predictions — select a behavioural profile that matches your subject. Each profile produces a distinct probability distribution based on how that subject type interacts with the environment.
  • Directional search — when the initial direction of travel is known, results are automatically focused to a directional cone, reducing noise and prioritising the most relevant area.
  • Weather-aware — the model factors in temperature, precipitation, and wind conditions. Weather is fetched automatically or can be overridden for historical analysis.
  • Flexible search areas — define your area as a circle, polygon, or buffered path. The API handles hex grid generation and clipping.

How to use it

  1. Choose a profile matching the subject’s behavioural category.
  2. Define a search area — circle (centre + radius), polygon (GeoJSON), or buffered path.
  3. Provide subject parameters — age, optional direction of travel, and date/time.
  4. The API returns a ranked list of hexagons with probabilities. Higher probability hexes should be searched first.

Endpoint

EndpointDescription
POST /v1/predictGenerate spatial predictions for a given profile and search area.

Area types

TypeFieldsMax size
circlecenter (lat/lng) + radius_m5 km radius
polygonGeoJSON Polygon geometry25 km²
pathGeoJSON LineString geometry + buffer_m500 m buffer

Weather

By default, the API fetches weather conditions automatically based on the search area centre and the provided datetime. You can:
  • Override weather — pass a weather object with temperature_c, precipitation_mmh, wind_speed_mps, and wind_direction_deg.
  • Disable weather — set disableWeather: true to use neutral defaults.

Confidence

The response includes a confidence level (high, medium, or low) indicating how well the model performs in the region of the search area.

Authentication

The Predict API uses the same x-api-key header as all other endpoints. See the authentication guide for details.