A comprehensive, AI-powered interactive study platform for CNT 4419: Secure Coding at the University of South Florida (Spring 2026). Everything runs entirely in your browser — no server, no data leaves your machine.
Built with Nuxt 3, Nuxt UI, Typst, and client-side AI.
Full, hand-typed lecture notes compiled from Typst to PDF and viewable directly in the browser with zoom, search, and page navigation. Covers all 8 chapters of the course:
- Foundations — Policy, mechanism, traces, programs
- Properties & Policy — CIA triad, safety & liveness, properties vs. non-properties
- Mechanisms — Static/dynamic, sound/complete/precise enforcers, false positives/negatives
- Mechanism Categories — Preventative, detective, containment, recovery
- Secure Design Principles — Input sanitization, error handling, defense in depth, PoLP, TOCTOU, complete mediation
- Attack Vectors — Trust categories, confused deputy, social engineering, phishing, DoS/DDoS
- Access Control — Authentication, authorization, ACL/capabilities/ACM, RBAC, MAC, DAC, Bell-LaPadula, Biba
- Memory & Buffers — Stack frames, buffer overflows, NOP sleds, return-to-libc, ROP, stack guards, ASLR, CFI, fuzzing
A full quiz system with over 500 pre-generated questions across all 8+ chapters, featuring:
- Multiple question types: Multiple choice (with easy/medium/hard difficulty), True/False, and Short Answer
- Chapter filtering: Focus on specific chapters or mix across all
- Definitions sub-chapter: A dedicated "Definitions Only" toggle that filters to 78 vocabulary and abbreviation questions covering every key term from the notes (CIA, PoLP, TOCTOU, ASLR, ROP, Bell-LaPadula, etc.)
- Self-grading for short answer: Submit your answer, compare against the key, and self-assess
- AI-powered explanations: When you get a question wrong, an on-device LLM explains why the correct answer is right, using context from the actual class notes
- XP & leveling system: Earn XP for every answer (more for correct ones) and level up from Rookie to Master
- Score breakdown: Per-chapter performance analysis after each quiz
- Confetti celebrations: Score 80%+ and get rewarded
Ask natural language questions about the course material:
- Semantic search (all browsers): Uses Transformers.js embeddings to find the most relevant note sections for any query
- Full conversational AI (Chrome/Edge with WebGPU): On-device LLM generates detailed answers grounded in the actual class notes via RAG (Retrieval-Augmented Generation)
- Privacy-first: Everything runs locally in your browser — no API calls, no data collection
Review key concepts in a flashcard format for spaced-repetition-style study.
- Dark/light mode with system preference detection
- Fully static deployment — no backend, just HTML/CSS/JS on GitHub Pages
- Responsive design — works on desktop, tablet, and mobile
- Offline-capable — once loaded, works without an internet connection
The latest PDF is automatically built and attached to a GitHub Release:
typst compile --features html cnt_4419_typed_notes.typThe --features html flag enables Typst's target() function, which the
chapter shim (notes/_html_shim.typ) uses to switch between the paged-PDF and
HTML renderings of admonition blocks. The flag does not change the PDF output;
it only unlocks the conditional API.
cd site
bun install
bun run dev# Generate embeddings from PDF
node scripts/extract-and-embed.mjs
# Add definition-based questions
node scripts/add-definitions-round.mjs
# Refine question bank with prompt variations
node scripts/refine-question-bank-round2.mjsNote on lecture content: earlier versions of this project shipped lecture recordings, auto-transcribed transcripts, and a "Professor Says" announcements view derived from those transcripts. Those features have been removed because publishing lecture recordings or transcripts requires the lecturer's written consent under Florida Statute § 1004.097 and USF Policy 10-048. Only the typed notes (the student's own work) and the question bank derived from them remain.
- Class notes are written in Typst (
cnt_4419_typed_notes.typ) — a modern typesetting system with LaTeX-quality output - On every push, GitHub Actions:
- Compiles the Typst source to PDF
- Extracts text and generates semantic embeddings for RAG search over the notes
- Builds and deploys the Nuxt static site to GitHub Pages
- Attaches the PDF to a GitHub Release
- The quiz engine draws from a curated bank of 543 questions across 8 chapters, with difficulty levels, multiple question types, and a definitions sub-chapter for vocabulary review
- The AI system uses Transformers.js for semantic search and WebLLM for chat — everything runs client-side with zero data exfiltration. The retrieval index is built only from the student-authored typed notes.
| Layer | Tool |
|---|---|
| Notes | Typst |
| Framework | Nuxt 3 |
| UI | Nuxt UI + Tailwind CSS |
| PDF Viewer | vue-pdf-embed |
| AI Search | Transformers.js (semantic embeddings) |
| AI Chat | WebLLM (on-device LLM via WebGPU) |
| Package Manager | Bun |
| Deployment | GitHub Pages (static, zero-cost) |
| Chapter | Questions | MC | T/F | SA | Definitions |
|---|---|---|---|---|---|
| 1. Foundations | 26 | 19 | 5 | 2 | 7 |
| 2. Properties & Policy | 32 | 25 | 5 | 2 | 10 |
| 3. Mechanisms | 31 | 23 | 5 | 3 | 8 |
| 4. Mechanism Categories | 26 | 20 | 4 | 2 | 8 |
| 5. Secure Design | 109 | 73 | 17 | 19 | 11 |
| 6. Attack Vectors | 82 | 58 | 7 | 17 | 5 |
| 7. Access Control | 117 | 94 | 19 | 4 | 12 |
| 8. Memory & Buffers | 120 | 99 | 14 | 7 | 17 |
| Total | 543 | 411 | 76 | 56 | 78 |