This is a submission for the Presage SmartSpectra SDK Challenge.

What I Built

I built HumanKit, a desktop app that uses the Presage SmartSpectra SDK to read body signals through a webcam and turn them into useful tools.

The idea behind HumanKit is simple: our bodies give off signals while we work, speak or try to slow down, but we do not always notice them. HumanKit helps people see some of these signals and use them in simple ways.

It has three tools:

1. SpeakEasy – Interview and Presentation Practice

SpeakEasy helps people practise for interviews and presentations. It asks practice questions, records spoken answers and gives feedback on the answers.

It also tracks breathing rate, pulse, speaking time, pauses and speaking speed. It compares breathing and pulse readings with the person's own resting baseline.

2. ScreenBreak – Blink Tracking and Eye Breaks

ScreenBreak counts how often a person blinks while working. It shows blink rate and reminds the person to take short eye breaks.

It also follows the 20-20-20 eye break habit.

3. BreatheBuddy – Guided Breathing

BreatheBuddy guides people through simple breathing exercises using an animated orb.

When the camera is enabled, it also shows the person's measured breathing rate and a live chest-movement waveform. This lets the person compare their breathing with the guide.

HumanKit is built with Electron, Next.js, React and TypeScript.

Demo

I could not include a demo video because of a camera issue.

The webcam available to me could not maintain the minimum 25 frames per second required by the Presage SmartSpectra SDK. Because of this, I could not get a reliable live reading to record.

I did not want to upload a video that could give the wrong impression of a live measurement. The code contains the Presage integration, but I was unable to record a working live demo on my available camera.

Code

View the HumanKit GitHub Repository

The repository includes the source code and a README with setup instructions.

How It Responds to Human Signals

HumanKit uses the Presage SmartSpectra Node.js SDK to read signals from the camera.

It uses these signals:

  • Breathing rate: Used in SpeakEasy to compare breathing during practice with a resting baseline. It is also used in BreatheBuddy to compare actual breathing with the guided breathing pace.
  • Pulse rate: Used in SpeakEasy to show changes from the person's resting baseline.
  • Blinking: Used in ScreenBreak to count blinks and show blink rate.
  • Talking and silence: Used in SpeakEasy to show speaking time and pauses.
  • Chest movement: Used in BreatheBuddy to show a live breathing waveform.

I connected the SDK through a shared file called lib/presage.ts. This file handles starting and stopping measurements, receiving readings and sending the results to the different tools.

Each tool only requests the signals it needs. For example, ScreenBreak only requests face signals for blink tracking, while BreatheBuddy requests breathing signals.

HumanKit also checks the quality of readings before using them. It does not simply display every number received.

The app gives simple instructions when the camera needs adjustment. For example, it can tell the user to improve the lighting, move closer or farther away, sit still, or bring their upper chest into view.

Prize Categories

I am entering Best Thoughtful Real-World Use.

HumanKit is built around three everyday situations: preparing for an interview or presentation, spending long hours looking at a screen, and taking time for guided breathing.

I wanted to make the signals useful instead of just showing numbers on a screen.

For example, someone practising for an interview can compare their breathing and pulse with their own resting baseline. Someone working at a computer can keep track of their blinks and get reminders to take eye breaks. Someone using BreatheBuddy can compare their actual breathing with a guided pace.

I also wanted the app to be easy to understand. It uses simple instructions, shows when a signal is unavailable, and asks for camera consent before using the camera.

Consent and Data Handling

HumanKit asks for consent before using the camera. The camera does not start until the user agrees and starts a session.

The microphone is only used in SpeakEasy when a person records an answer.

Camera and video

HumanKit does not record or save video. The camera stream is passed to the Presage SmartSpectra SDK for processing. The camera stops when the session ends or the user leaves the tool.

Data sent to Presage

The camera stream is passed to Presage. The SDK also needs an internet connection to check the API key and provide its service.

Presage's optional usage telemetry is disabled in HumanKit. This does not mean that all network communication from the SDK is disabled.

Data sent to Groq

SpeakEasy uses Groq for AI features. Groq receives the practice topic, recorded answer audio for transcription, and answer text with selected measurements for feedback.

ScreenBreak and BreatheBuddy do not use Groq.

Data saved on the computer

HumanKit saves short session summaries and the user's camera consent choice in local storage on the device.

It does not save video, recorded audio or the full text of spoken answers in session history.

There are no HumanKit accounts or HumanKit servers.

Data deletion

Users can delete all locally saved data from the Privacy page. They can also delete individual sessions from History.

This only deletes data saved by HumanKit on the device. It does not delete information already sent to Presage or Groq. HumanKit does not control how long those services keep information they receive.

API keys

For this version, the Presage and Groq API keys are supplied through the local environment file. The current build setup includes these keys in the built app, so users should not share an installer containing their personal keys.

Health and safety

HumanKit does not diagnose, treat or give medical advice. Its measurements are estimates. They are not proof of stress, confidence, health or how someone feels.