BreatheAI Presage SmartSpectra SDK Challenge

This is a submission for the Presage SmartSpectra SDK Challenge

What I Built

BreatheFlow AI — A Voice-First Adaptive Breathing Companion

BreatheFlow AI is a camera-assisted, voice-first wellness companion that helps people take a moment to reset, whether they are feeling stressed before an exam, tired after a long meeting, or simply looking for a moment of calm.

The experience begins with a short camera-based wellness check-in powered by the Presage SmartSpectra SDK. The SDK analyzes supported signals and produces a baseline report. Users can then share how they are feeling in their own words.

BreatheFlow combines the available SmartSpectra report with the user's personal context and uses Gemini to generate a personalized breathing recommendation. Gnani AI provides conversational voice interaction and spoken guidance, while an adaptive breathing pacer guides the user through the session.

The goal is to make breathing exercises feel less generic and more responsive to the person using them.

Demo

Demo video: https://drive.google.com/file/d/1Vdi85afushRgbi-L_lBjS2RJv1j0Ccea/view?usp=sharing

The demo will show the complete experience:

1. Starting a camera-based wellness check-in.
2. Capturing live signals through Presage SmartSpectra.
3. Generating and displaying the wellness report.
4. Sharing personal context through voice or text.
5. Generating a personalized breathing protocol with Gemini.
6. Following the guided breathing session with the adaptive pacer and spoken guidance.

Code

GitHub Repository: BreatheFlow AI

The repository contains the application source code and setup instructions.

The project uses React, TypeScript, Electron, FastAPI, Gemini, Gnani AI, and the Presage SmartSpectra native SDK.

How It Responds to Human Signals

BreatheFlow uses Presage SmartSpectra to obtain camera-derived wellness signals, including supported respiration and pulse measurements.

The integration follows a short baseline check-in:

1. The user explicitly starts camera sensing.
2. The SmartSpectra SDK processes the camera stream and performs its supported signal analysis.
3. The application collects the metrics and statuses actually returned by the SDK.
4. A wellness report is generated and shown to the user.
5. The report, together with the user's own description of their current situation, is sent to Gemini.
6. Gemini generates a personalized breathing protocol, which the application validates before passing it to the breathing pacer.

For example, a user might say, "I'm stressed about my exam." BreatheFlow considers that context alongside the available SmartSpectra report and may recommend a gentle, structured breathing exercise to help the user take a moment to relax and regain focus.

The breathing pacer executes the protocol using a deterministic timer, while Gnani AI provides spoken guidance.

The SDK report and the user's self-reported feelings remain distinct inputs. BreatheFlow does not treat pulse or respiration measurements alone as proof of a particular emotional state.

Prize Categories

Best Thoughtful Real-World Use

BreatheFlow AI is designed for people who experience everyday stress, mental fatigue, or moments when they need to pause and reset.

Breathing exercises are widely accessible, but many applications provide the same fixed routine regardless of the user's situation. BreatheFlow explores a more personalized approach by combining camera-derived wellness signals with the user's own description of how they feel.

The experience is designed around a simple interaction: check in, understand the available signals, share your context, and receive guided breathing support.

The camera is an optional part of the experience. Users can still access voice-based breathing guidance when native camera sensing is unavailable.

The project is intended for general wellness and mindfulness—not medical diagnosis, treatment, or clinical decision-making.

Consent and Data Handling

BreatheFlow requires the user to explicitly initiate camera sensing. Camera access is not intended to run covertly or without the user's knowledge.

The SmartSpectra SDK processes the camera stream to obtain its supported wellness signals. The application uses the resulting report to personalize the breathing experience.

The report and the user's voluntarily provided context are sent to Gemini for protocol generation. Gnani AI is used for conversational voice interaction and spoken guidance.

The application also supports user authentication and session persistence. Historical sessions are not deleted when a user starts a new check-in.

BreatheFlow does not claim to provide medical diagnoses or make medical decisions based on camera-derived measurements.