MindHub — Privacy-Aware Mental Health Workflow Demo

About this work

Demo and research prototype that grew out of an AI Reboot Academy team MVP. It records, summarises and hands off — it does not diagnose. Every patient in it is fictional.

Challenge

Patients remember the week in fragments and a consultation gets fifteen minutes. The signals a clinician wants — sleep, mood, medication, stressors — sit scattered across daily notes nobody reads. Gathering them is only acceptable if the patient decides exactly what crosses over.

Approach

Daily notes become a pre-visit summary bounded by a sharing scope the patient sets. Roles are separated in the database, not only in the interface — a patient account cannot escalate itself to doctor, and fields marked private are never sent to the doctor browser. Risk-keyword detection runs rule-based on the front end, so it still works when the AI call fails. The AI proxy enforces allowed origins, request-size and input-length caps, per-IP rate limits and upstream timeouts.

Verification

Three entry points are checked in real browsers — the no-login interactive demo, the patient app and the read-only doctor report — at desktop, tablet and mobile widths. The role-escalation and sharing-scope policies live in the repository as SQL rather than as a claim in a slide.

Live demo — https://mindhub.forblune.com