Raw Features

Raw features are direct representations of sensor and activity data as collected by the mindLAMP platform. They form the base layer of Cortex's three-tier pipeline.

What Raw Features Do

Raw features query the LAMP API and normalize low-level data streams into a consistent format for downstream computation. They handle iOS/Android differences automatically — see the Features overview for details on the full pipeline.

Complete Raw Feature List

Sensor Raw Features (11)

Raw Feature	Source Sensor	Consumed By	Description
gps	lamp.gps	significant_locations, trips, data_quality	Latitude, longitude, altitude, accuracy
accelerometer	lamp.accelerometer	acc_jerk, data_quality	x, y, z acceleration (g)
device_motion	lamp.device_motion	—	Rotation, gravity, magnetic field, attitude
device_state	lamp.device_state	screen_active, battery_level	Screen on/off, battery level
steps	lamp.steps	step_count	Step count and type
nearby_device	lamp.nearby_device	nearby_device_count	Bluetooth/WiFi detections
telephony	lamp.telephony	call_number, call_duration, call_degree	Call metadata
analytics	lamp.analytics	step_count (version detection), app_time	App usage events
sleep	lamp.sleep	healthkit_sleep_duration	Sleep stage data
screen_state	lamp.screen_state (deprecated)	inactive_duration	Legacy screen state events
ambient_light	—	—	Ambient light sensor (limited availability)

SensorKit Raw Features (5)

These features require iOS SensorKit data, available only on enrolled research devices.

Raw Feature	Description	Consumed By
device_usage	Screen time, unlock events, app usage	screen_unlocks, screen_unlock_duration, screen_wakes, app_time
phone_usage	Phone call metadata (SensorKit source)	call_number, call_duration, call_degree (SensorKit mode)
messages_usage	Text message metadata	text_number, text_degree
visits	Location visit data with categories	visit_time

Activity Raw Features (13)

Raw Feature	Source Activity	Consumed By	Description
survey	lamp.survey	survey_scores	Question-level survey responses
jewels_a	lamp.jewels_a	game_level_scores	Jewels A game events
jewels_b	lamp.jewels_b	game_level_scores	Jewels B game events
balloon_risk	lamp.balloon_risk	game_level_scores	Balloon Risk events
cats_and_dogs	lamp.cats_and_dogs	game_level_scores	Cats and Dogs events
pop_the_bubbles	lamp.pop_the_bubbles	game_level_scores	Pop The Bubbles events
spatial_span	lamp.spatial_span	game_level_scores	Spatial Span events
dcog	lamp.dcog	—	D-Cog events
digit_span	lamp.digit_span	—	Digit Span events
fragmented_letters	lamp.fragmented_letters	—	Fragmented Letters events
funny_memory	lamp.funny_memory	—	Funny Memory events
trails_b	lamp.trails_b	—	Trails B events
voice_survey	lamp.recording	—	Voice recording audio data

Legacy Raw Features

The following raw features correspond to deprecated sensors. They remain in Cortex for backward compatibility with older data. For the full deprecated sensor migration guide, see Deprecated Sensors.

Raw Feature	Deprecated Sensor	Replaced By
bluetooth	lamp.bluetooth	nearby_device
calls	lamp.calls	telephony
sms	lamp.sms	telephony
gyroscope	lamp.gyroscope	device_motion

Usage

Raw features are typically not called directly — they are consumed by primary and secondary features automatically through Cortex's dependency graph. However, they can be accessed directly for custom analysis:

import cortex

# Access raw GPS data
gps_data = cortex.raw.gps(id="U1234567890", start=start_time, end=end_time)

# Access raw survey data
survey_data = cortex.raw.survey(id="U1234567890", start=start_time, end=end_time)

Output Structure

All raw features return a consistent structure:

{
    "data": [ ... ],        # List of sensor/activity event records
    "has_raw_data": 1       # 1 if data exists, 0 if no data found
}

The has_raw_data flag distinguishes between "no data was collected" (0) and "data was collected but may be empty after filtering" (1).