Quickstart

Go from zero to a scored voice sample in four steps: install the SDK, register a patient, submit audio, and read the drift against their baseline.

1 · Install

# Node.js
npm install voxtar
# Python
pip install voxtar

2 · Register a patient

Patients are the anchor for longitudinal baselines. Use a pseudonymous external_id from your own system — no PHI required.

import { Voxtar } from "voxtar";
const voxtar = new Voxtar({ apiKey: process.env.VOXTAR_API_KEY });
const patient = await voxtar.patients.create({
  external_id: "subject-1042",
  age_band: "40-64",
});

3 · Analyze a voice sample

Submit 3–60 seconds of speech. A sustained vowel or short read passage gives the cleanest features.

import { readFileSync } from "node:fs";
const analysis = await voxtar.analyze({
  patient_id: patient.id,
  audio: readFileSync("sample.wav"),
  panels: ["respiratory", "neurological"],
});

4 · Read the result

The first few samples establish a baseline; after that, every analysis is scored against the patient's own norm and raises a flag when it drifts.

for (const flag of analysis.flags) {
  console.log(flag.panel, flag.code, "->", flag.severity);
}
// respiratory respiratory_drift -> review

Next: wire up Webhooks so drift alerts reach your care team in real time, or turn on FHIR write-back to land biomarkers straight in the EMR.