メインコンテンツまでスキップ

Knowledge

MapGPT taps into Mapbox' proprietary map data layers, which encompasses essential geographic information such as road networks, points of interest, addresses, and real-time traffic data. These data layers serve as the foundation for MapGPT's location-aware interactions, ensuring accurate and contextually relevant responses.

Suppose a user asks, "Where's the nearest station?", "How's the traffic ahead?", or "What can you tell me about this city?". In such cases, the AI assistant adeptly determines if the user is seeking a charging or fuel station, offers current traffic updates, or unveils hidden facets of a city you thought you knew inside out. As well as provide interesting facts or stories as a source of conversation, entertainment, and utility.

What is sent to MapGPT?

The data sent to MapGPT can be categorized into five main contexts: App, User, Vehicle, Route, and EV. Each of these contexts contains specific details that aid MapGPT in understanding and efficiently processing requests. This data is notably adaptable, allowing customization and expansion based on customer requirements. Contact sales for information.

Full context overview

Required

PropertyTypeDescription
userContextobjectInformation about the user's current location, point of interest, and prevailing weather conditions.

Optional

If not provided, the value will be assigned as null.

PropertyTypeDescription
appContextobjectDetails about the application, such as localization preferences, battery status, and more.
vehicleContextobjectSpecifics about the user's vehicle including type, battery status, and current conditions within the vehicle.
routeContextobjectData about the active navigation route, such as ETA, distance remaining, and potential alternate routes.
evContextobjectDetails about electric vehicles, primarily focused on nearby charging stations.
val context = MapGptContextDTO(
userContext = MapGptUserContextDTO(...),
appContext = MapGptAppContextDTO(...),
vehicleContext = MapGptVehicleContextDTO(...),
routeContext = MapGptRouteContextDTO(...),
evContext = MapGptEVContextDTO(...),
)

User context

Required

PropertyTypeDescription
latstringLatitude used for reverse geocoding, and searching nearby places.
lonstringLongitude used for reverse geocoding, and searching nearby places.
placeNamestringRecognizable name or title of the user's location or destination.
val userContext = MapGptUserContextDTO(
lat = "37.7694777",
lon = "-122.486135",
placeName = "Golden Gate Park, San Francisco, CA",
)

App context

Optional

If not provided, the value will be assigned as null.

PropertyTypeDescription
localestringThe language and regional settings of the application.
temperatureUnitsstringMeasurement units (e.g., Celsius or Fahrenheit) used for temperature display.
distanceUnitsstringMeasurement units (e.g., Kilometers or Miles) used for distance display.
clientTimestringCurrent time from the application's perspective.
mediaobjectActive media context, like now playing song or media state.
val locale = Locale.getDefault()
val media = MapGptMusicContextDTO(
name = "Shake It Off",
artist = "Taylor Swift",
state = "Playing"
)
val appContext = MapGptAppContextDTO(
locale = locale.toLanguageTag(),
temperatureUnits = "Fahrenheit", // or "Celsius",
distanceUnits = "mi", // or "km",
clientTime = SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss", Locale.getDefault())
.format(Calendar.getInstance().time),
)
media = media,

Vehicle Context

Optional

If not provided, the value will be assigned as null.

PropertyTypeDescription
fuelstringType of fuel or energy source the vehicle uses. Examples are gasoline, diesel, biodiesel, electric, hydrogen, and hybrid.
batteryLevelintegerFor electric and hybrid vehicles, the percentage of battery remaining.
acTemperatureintegerThe set temperature of the vehicle.
acStatusstringActive status of the air conditioning system (e.g., on, off).
val vehicleContext = MapGptVehicleContextDTO(
fuel = "Gas",
batteryLevel = 25,
acTemperature = 70f,
acStatus = "on",
)

Route Context

Optional

If not provided, the value will be assigned as null.

PropertyTypeDescription
activebooleanIndicates when the navigation experience is following turn-by-turn guidance.
etastringPredicted time of arrival to the current waypoint.
timeLeftintegerDuration remaining to reach the route destination in minutes.
distanceLeftdoubleDistance remaining to reach the route destination. The unit will be either in mi or km depending on appContext.distanceUnits
maxCongestionintegerThe level of congestion in numeric form, from 0-100. A value of 0 indicates no congestion, a value of 100 indicates maximum congestion.
typicalDurationintegerUsual travel time for the current route under normal conditions in minutes.
alternateRoutesintegerAvailable alternative routes to the current destination.
speeddoubleCurrent speed of the vehicle in meters per second.
speedLimitdoubleMaximum permissible speed for the current road segment in meters per second.
originobjectStart point of the current route.
destinationobjectEnd point of the current route.
val origin = MapGptLocationDTO(
lat = "37.7694777",
lon = "-122.486135",
placeName = "Golden Gate Park, San Francisco, CA",
)
val destination = MapGptLocationDTO(
lat = "37.8255",
lon = "-122.4994",
placeName = "Hawk Hill, San Francisco, CA",
)
val routeContext = MapGptRouteContextDTO(
active = true,
eta = "2024-09-09T16:16:00",
timeLeft = 19,
distanceLeft = 9.050983807299744,
maxCongestion = 38,
typicalDuration = 18,
alternateRoutes = 2,
speed = 8.673388481140137,
speedLimit = 11.176,
origin = origin,
destination = destination,
)

EV Context

Optional

If not provided, the value will be assigned as null.

PropertyTypeDescription
nearbyStationslistA list of nearby charging stations and their information, useful for electric vehicles.
val evStation1 = MapGptEVStationDTO(
name = "EV Station 1",
distance = "5 mi", // 8 km depending on distanceUnits from appContext
)
val evStation2 = MapGptEVStationDTO(
name = "EV Station 2",
distance = "6 mi", // 9.6 km depending on distanceUnits from appContext
)
val evContext = MapGptEVContextDTO(
nearbyStations = listOf(evStation1, evStation2)
)

When is it sent to MapGPT?

The context is a part of MapGptService.postPromptsForStreaming API call. Whenever a user issues a prompt, all the context mentioned above can be transmitted via the API. Out of all context only userContext is required and all others are optional. It is recommended to send the context if data is available, as sharing all with the AI system can significantly enhance the relevance and quality of interactions.

val context = MapGptContextDTO(
userContext = MapGptUserContextDTO(
lat = "37.7694777",
lon = "-122.486135",
placeName = "Golden Gate Park, San Francisco, CA",
),
appContext = MapGptAppContextDTO(
locale = "en",
temperatureUnits = "Celsius", // or "Fahrenheit",
distanceUnits = "km", // or "mi",
clientTime = SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss", Locale.getDefault())
.format(Calendar.getInstance().time),
),
)
val request = MapGptStreamingRequest(
prompt = prompt,
context = context,
)
mapGptService.postPromptsForStreaming(request)
MapGPT context with Dash
When using Dash application instead of MapGPT SDK, you don't need to construct the context as shown above. Dash application already constructs the context and sends it via the postPromptsForStreaming API call.

How to update the context with UX Framework?

When using MapGPT with UX Framework, the context is created with default values. There might be use cases where you would need to update the context with new data values. MapGPT provides a context override API that allows you to update the context at runtime. For example, if you want to update MapGPT about the car battery level being low at 10%, you can do the following:

Dash.controller.updateMapGptContextOverrides {
vehicleContext.value = { contextDTO ->
contextDTO.copy(
batteryLevel = 10
)
}
// Similarly, other contexts can be overridden
evContext.value = { contextDTO -> contextDTO.copy(...) }
appContext.value = { contextDTO -> contextDTO.copy(...) }
userContext.value = { contextDTO -> contextDTO.copy(...) }
routeContext.value = { contextDTO -> contextDTO.copy(...) }
}
この{Type}は役に立ちましたか?