
How Qwizflow Works: Architecture, Journeys & Security
By Qwizflow Team
17 April 2026
· 11 min read
Qwizflow is not just a quiz app — it is a multi-layered AI learning platform serving four distinct roles with enterprise-grade security, with learning records stored in Australia. This guide walks through how every piece fits together.
1. Platform Architecture
Every interaction follows the same secure pipeline — from the button you tap to the AI that responds.
- Your Dashboard — the React app in your browser
- Secure Gateway — auth token plus consent check
- Learning Engine — business logic and mastery scoring
- Database — Melbourne, Australia
- AI Providers — prompts sent with email addresses and phone numbers masked by pattern
The whole path is TLS 1.3 encrypted, rate-limited to 100 requests per minute per user, protected by a 7-day recovery window, and personal details such as email addresses and phone numbers are masked by pattern before a prompt reaches AI — a pattern cannot catch everything a learner types, so we ask learners not to type personal details.
1b. The Owl Tutor — One Chat Front Door
Layer 1 used to mean "pick the right AI tool from a menu". With the Owl Tutor that step is gone — students describe what they want in plain language and the right tool just happens inside the same chat.
- You Ask — type (or, from 13, speak) into the Owl chat
- Owl Reads Intent — decides which tool fits
- Same Consent and Caps — the per-feature gate still applies
- Tool Runs In-Chat — cards, quizzes, voice
- It Remembers — cross-session memory, 90-day history
The chat is grounded in the same single, living learner picture every other surface reads from — content level, interests, outstanding misconceptions, recent topic summaries, durable notes from parents and teachers — so whatever tool the Owl spins up sounds like "the same tutor" instead of starting from scratch. If your ask needs a feature your parent hasn't approved yet, the chat shows a one-tap bundle approval right there instead of dead-ending you. Conversations are saved for 90 days so you can pick up where you left off; older history quietly auto-expires. Replies read aloud, voice typing from 13, camera-to-chat, per-feature consent unchanged.
2. Feature Map by Role
Every user sees a different experience. Parents get progress tracking, AI insights, the family hub, consent controls and notes on their child's learner profile. Teachers get classroom management, analytics, teaching tools, AI assistants including lesson and unit planners, the help queue, the affective heatmap, safety flags and messaging. School admins get the roster, school-wide AI policy, the school library and school paths, compliance checks and outcomes reporting.
Students find their tools in the hub, grouped into zones:
- Today — assignments and what to do next
- Learn — learning paths, knowledge maps, study materials and the school library
- Practice — quizzes, review, voice quizzes, challenges and Focus Mode
- AI Studio — Socratic Tutor, Homework Helper, Explain Answer, AI flashcards, mnemonics, Pocket Lecture, study guides, Story Writer, Comic Explainer, Remix Studio, audio podcasts, translation, and roleplay and debate from 16
- Compete — live quizzes, standings and studying together
- Grow — progress, goals and streaks, your portfolio and achievements
The Unified Learner Picture
Every student on Qwizflow has a single, living learner picture that the whole platform shares. When the AI Tutor answers a question, the Story Writer continues a chapter, or the Debate Arena responds to an argument, it is all reading from the same profile — so you get a consistent experience instead of starting from scratch every time. It holds:
- Mastery Rollup — strengths, struggles, recent topic summaries
- Reading Level — per-subject target reading level plus confidence
- Pace Signature — how long you study and when you peak
- Affective Rollup — confidence, frustration, mood trend
- Mode Weights — visual / audio / interactive / text preferences
- Goals — targets with countdowns and linked subjects
- Error Taxonomy — slip / procedural / conceptual / vocab / prereq
- Tone Preference — too dry / just right / too babyish signal
- Annotations — durable notes from parents and teachers
Plus a rolling 60-minute working memory that remembers what you just did across features, so the AI can hand off naturally when you jump from a quiz to a tutor to a story. You see your own learner picture on your profile. A linked parent can see their child's. Teachers do not read the raw picture — they see aggregated classroom analytics. Email addresses and phone numbers, and names written behind a label, are masked by pattern before a prompt leaves our servers.
3. Student Journey: Quiz and Mastery
- Choose Your Curriculum — pick the Australian Curriculum or the US K–10 catalogue (Common Core for English and maths, NGSS* for science), choose your year level and subjects, and add ready-made topics.
- AI Generates Your Quiz — an adaptive quiz calibrated to your level; a consent check ensures you have opted in to AI quiz generation, with an inline prompt if not.
- Answer with Live Feedback — a live mastery bar tracks your estimated score in real time; use progressive hints or request an easier question variant when stuck.
- Mastery Score Calculated — answers are scored across 5 weighted dimensions: Accuracy 50%, Calibration 20%, Concepts 15%, Effort 10%, Retention 5%.
- Three-Way Decision — score 0.75 or above passes the topic and earns a trophy; 0.5–0.74 goes to remediation with a targeted study guide and up to 2 retry attempts; below 0.5 the topic is blocked and the AI generates a misconception lesson.
- Earn Trophies and Progress — bronze, silver, gold and platinum trophies; the path updates, the next topic unlocks, and spaced repetition schedules future reviews.
4. Student Journey: Misconception Remediation
- Topic Blocked — a Misconception Alert banner explains which concept is blocking progress.
- AI Generates a Personalised Lesson — targeted at your exact misconception, with worked examples and visual explanations.
- 3-Question Retest — server-graded, with a 30-minute session limit and a maximum of 3 sessions per hour.
- Misconception Resolved, Topic Unblocked — pass and your mastery updates; fail and the AI generates a new lesson targeting where you went wrong this time.
5. Parent Journey: Monitor and Support
- View Linked Children — each card shows streak, active subjects, and a mastery snapshot.
- Review Progress and Weekly Digest — mastery per subject, quiz completion, time spent, week-over-week trends, and a plain-language digest.
- Get AI-Powered Insights — the Enhanced Digest identifies patterns, Help Tonight suggests dinner-table conversation starters, and the AI Coach answers questions about your child's learning.
- Manage AI Consent and Review Transparency — toggle individual features in the AI Consent Centre, with age-appropriate "Explain this toggle" tooltips, a Recent Activity view (last 10 interactions per feature), Transparency Cards, a timestamped Consent Audit Log, and the AI Transparency Ledger with one-click erase and JSON export.
- Family Hub and Caregivers — invite co-parents, grandparents, or tutors with per-caregiver permissions; when two parents are linked, the child has one consent record — the latest change by either parent applies, and every change is kept in a consent history both can review.
6. Teacher Journey: Classroom Command Centre
- Create and Manage Classrooms — unique join codes, CSV bulk import, classroom-level AI policies and active hours.
- View Student Analytics — the Knowledge Heatmap shows mastered (green), learning (amber), and not started (grey) concepts, with drill-down student reports.
- Progress Alerts Flag Who Needs a Check-In — fixed, visible rules (three quiz scores in a row below 50%, no activity for five days, or falling mastery) raise an alert that shows its reason, only to that class's teacher; no AI model decides who is flagged.
- Launch a Live Quiz — Kahoot-style with join codes, a real-time leaderboard, and exportable results.
- Monitor Students Live — see who is answering, who is idle, who just had a breakthrough; one-click nudge; the AI Monitor shows aggregated AI usage patterns.
7. School Admin Journey: Governance and Oversight
- School Overview — students, teachers, classrooms, engagement, AI adoption, compliance status.
- Approve Teachers — teachers register with a secure invite code (TCH-xxxxx) and wait in your approval queue.
- Set School-Wide Policies — policies cascade to all classrooms; teachers can further restrict but cannot override you to enable a feature you disabled.
- Check Compliance and Report Outcomes — a compliance check covering consent records, the safeguarding counsellor, pending teacher approvals, ST4S registration and the school's AI content source, plus an outcomes view with a board-ready PDF.
8. Consent Architecture
Age tier determines who manages consent: under 13, AI is blocked until a parent consents (parent only); ages 13–15, AI is available if a parent approves (parent only); ages 16–17, the student can self-consent with parental override; 18+ has full autonomy.
Consent is granular per feature — quiz generation, help after a wrong answer, quiz pictures, adaptive tests, the diagnostic, Homework Helper, Socratic Tutor, the Owl chat, study guides, Story Writer, creative tools, Debate Arena, roleplay, translation, religious studies content, help requests to a teacher, the tutor unlock interview and adaptive learning, with separate switches for the grown-ups' own AI tools — and enforced by a two-sided gate: the frontend checks consent before showing any AI UI (an inline prompt appears if consent is missing), and every AI request is re-checked server-side, which cannot be bypassed from the browser. Multi-parent rule: one consent record per child, the latest change applies, and every change is kept in the consent history.
9. Role-Based Access Control (RBAC)
Authentication flows from Google Sign-In (OAuth 2.0) to a signed, short-lived JWT, a server-assigned role check, and finally role-scoped data access. Students see their own data only (invite code CLS-xxxxx). Parents see linked children only — progress summaries, consent settings, AI insights, never raw transcripts (PAR-xxxxx). Teachers see their own classrooms only — aggregated analytics and safety flags, never raw AI conversations (TCH-xxxxx). School admins get school-wide oversight (isSchoolAdmin flag). Ownership is enforced: no horizontal privilege escalation.
The Talking 3D AI Character Pipeline
Every voice surface — Live Tutor, Discussion Mode, Read Aloud — is fronted by a 3D character that talks as the audio plays. How a spoken word reaches the screen: the AI composes a reply anchored to the learner profile, Google Cloud Text-to-Speech renders it in the included read-aloud voice, the character's mouth shapes are generated from the words and timed to the audio, and the seconds played come out of the learner's daily read-aloud allowance. There are five voice personas to choose from, and a reduced-motion path for anyone who turns animation down.
Your AI Allowance — You Can See It
Every learner gets a free AI allowance each month — enough for about 40 new quizzes — that starts again at the beginning of each month, plus a separate daily allowance of read-aloud time, which is smallest for the youngest learners. A battery chip in the student hub shows what is left, with a breakdown for text, pictures and voice. Qwizflow sets these allowances; a school does not edit them. Past the free allowance, a parent can add an optional one-off credit pack, and a learner can buy one for themselves from 16.
10. Platform at a Glance
Four roles — student, parent, teacher and school admin — with per-feature AI consent, a free AI allowance each month, and learning records stored in Australia. TLS 1.3 encryption, a 7-day recovery window, daily auto backups, a 100-requests-per-minute rate limit, and personal details masked by pattern before AI calls.
FAQ
Is Qwizflow free to use?
Yes — Qwizflow is free to start, with a free AI allowance each month (about 40 new quizzes) across the student, parent, teacher and school admin roles. Credit packs are optional and one-off; there is no subscription.
Where is Qwizflow student data stored?
Learning records are stored in Australia — the database sits in Melbourne — protected by TLS 1.3 encryption, daily automated backups and a 7-day point-in-time recovery window. Sign-in runs on Firebase Authentication in the United States, and AI requests go to Google's global endpoints. Personal details are masked by pattern before a prompt reaches an AI provider.
How does Qwizflow decide when a topic is mastered?
Quiz answers are scored across five weighted dimensions — accuracy (50%), calibration (20%), concepts (15%), effort (10%) and retention (5%). A score of 0.75 or above passes the topic, 0.5–0.74 triggers remediation with up to 2 retries, and below 0.5 generates a targeted misconception lesson with a 3-question retest.
Who manages AI consent for a child on Qwizflow?
It depends on age: under-13 accounts have AI blocked until a parent consents, ages 13–15 need parental approval, ages 16–17 can self-consent with parental override, and 18+ has full autonomy. Consent is per-feature and enforced on both the frontend and the backend.
What happens when a student's free AI allowance runs out?
New AI content waits until the allowance starts again at the beginning of the next month, unless a parent adds an optional credit pack or, from 16, the learner buys one. The battery chip in the hub shows what is left before that happens. Read-aloud time is a separate daily allowance.