Skip to content

RorriMaesu/Gnosys-AI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

224 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Gnosys-AI Logo

Gnosys-AI 🧠 (Universal Study Suite)

Deploy to GitHub Pages HTML5 JavaScript TailwindCSS Ollama

Buy Me A Coffee

Gnosys-AI is a premium, AI-powered educational suite designed for students across all disciplines. It accelerates academic mastery through highly gamified spaced-repetition, intelligent scenario decryption, scientific calculations, and an integrated local LLM tutor powered by Ollama with configurable model choices.

Note

Author's Note & Disclaimer: I am a student navigating this learning path and developed this application for my personal study. While I am happy to share this tool with other students, please note that it is provided "as-is" with absolutely no guarantees of accuracy. I am using it to learn, and encourage you to always cross-reference definitions with official textbooks.

Gnosys-AI Hub Landing Page


🌟 Features

1. The Study Hub

A sleek, modern central landing page providing access to the complete curriculum required for academic and professional success. The entire suite features fully integrated Light and Dark Mode options with beautiful, frosted-glass navigation and fluid theme switching.

Gnosys-AI Landing Page Focus Pomodoro Timer

2. Medical Terminology (Active Module)

The foundational module for clinical language mastery.

  • Spaced Repetition Flashcards: Adaptive algorithm to ensure long-term retention of medical roots, prefixes, and suffixes.
  • Word Builder: Drag and drop prefixes, roots, and suffixes to construct complex clinical terms.
  • Chart Decrypter: Translates complex clinical SOAP notes into plain English, intelligently graded by an LLM.
  • Abbreviation Decoder & Pluralization: Rapid-fire tools to master medical shorthand and grammar.
  • Dr. Lex (AI Tutor): An integrated, warm, and precise AI tutor powered by your selected local Ollama model (including Gemma, Llama, Qwen, Mistral, Phi, or custom tags) that can break down complex terms, provide mnemonics, and generate on-demand quizzes.
Flashcards Word Builder
Chart Decrypter Dr. Lex AI Tutor
Abbreviation Decoder Pluralization
Lexicon Dictionary

3. Intro to Chemistry (Active Expansion)

A full chemistry learning system spanning Skills Lab, Coursework, and Homework Binder, built for CH 104Z/124Z-style progression.

  • Skills Lab + Crash Course Hub: Dedicated Crash Course Chemistry section with episode playlist navigation, watched tracking, and micro-quiz reinforcement.
  • Core Calculator Stack: Molar mass, dimensional analysis, stoichiometry, significant figures, nomenclature, and lab-focused precision tools.
  • Prof. Beaker (AI Tutor): Context-aware chemistry guidance powered by your local Ollama model.
Chemistry Dashboard Coursework Lectures
Periodic Table Prof. Beaker AI Tutor
Dimensional Analysis Measurement Lab

4. Applied Mathematics (Active Module)

A mathematical engine designed to reinforce safety, precision, and accuracy in scientific and clinical calculations.

  • Numeracy & Units: Practice converting between SI metrics, legacy units, concentrations, and physical measurements.
  • Precision Validator: Catch and correct formatting errors like trailing zeros (5.0), naked decimals (5.), and missing leading zeros (.5).
  • Algebraic Logic: Master formula rearrangement with an interactive, foolproof "variable chip" puzzle UI.
Clinical Numeracy Safety Validator
Algebraic Logic Chips Math Tutor Chat

5. Intro to Psychology (PSY 201Z) (Active Module)

A full psychology learning system spanning dashboard, Crash Course, coursework, and homework mastery flow.

  • Dashboard + Navigation Parity: Chemistry-style dashboard with desktop/mobile section navigation, quick links to coursework/homework, progress cards, and integrated stats modal.
  • Crash Course Psychology Path: A dedicated 41-episode learning track with watched-state tracking, progression CTA, and concept-targeted 3-question micro-quizzes powered by PsychTutor.
  • Interactive Skills Lab Tabs: Structured practice across Research Design, Axon Signaling, Conditioning, Memory Systems, and Clinical Intake to reinforce core intro psychology competencies.
  • Coursework Portal: Lesson-driven progression with lecture flow, Socratic checkpoints, sandbox verification, and Feynman explanation stages.
  • Homework Binder: Gate-aware worksheet queue with mastery scoring (>=80%), lesson-state progression, and alignment to coursework unlocks.
  • PsychTutor + Gamification: Embedded tutor support throughout activities plus XP, levels, streaks, daily quests, and achievement tracking linked to module behavior.
Psychology Dashboard Psychology Crash Course
Psychology Coursework Psychology Homework Binder

6. Gamification Engine

Learning complex clinical and academic data is hard. Gnosys-AI keeps you engaged and addicted to studying.

  • XP & Leveling System: Progress from Pre-Med Student all the way to Legacy Master.
  • Streak Tracking & Shields: Maintain daily study streaks, unlock achievements, and earn shields to protect progress.
  • Achievement Badges: Unlock over 20 unique badges.
  • Daily Quests: Complete objectives for bonus XP.
Stats Dashboard Level Up

7. Fully Responsive Mobile UI & On-Device AI Engine

Study anywhere, perfectly optimized for your phone with our dedicated on-device local AI support:

  • On-Device Gemma 4 Execution: Bypasses local PC server limitations! Runs Google's Gemma 4 model directly in your browser tab using WebGPU acceleration via the high-performance LiteRT-LM (formerly TFLite GenAI) framework.
  • Hardware-Aware Model Recommendations: Automatically detects the smartphone's physical memory using navigator.deviceMemory:
    • Recommends and pre-selects the Gemma 4 Efficient (2.5B) tier (~2.01 GB) on devices with < 8GB RAM to prevent Out-Of-Memory (OOM) browser tab crashes.
    • Recommends and pre-selects the Gemma 4 Pro (4.5B) tier (~2.97 GB) on flagship devices with 8GB+ RAM.
    • Allows full manual overrides so users remain in control of their hardware configuration.
  • Offline No-AI Fallback: An option to "Continue without Local AI" saves mobile resources and local space (~2 GB OPFS file bypass). If the student chats with a tutor while offline, Gnosys-AI streams a friendly, zero-overhead tutorial card explaining how to re-activate it.
  • Universal Header Toggles: The top status badge of every single page is styled with hover transition states and pointer cursors, acting as a global toggle. Click the badge at any time to open the configuration modal.
  • Preface History Batch Prefilling: Feeds the entire conversational context (system prompt + history exchanges) in a single, parallel preface batch pass. Prefill speeds improve by 10x-20x (under 1.5 seconds on phone GPU/CPUs), starting the tutor token stream instantly instead of hanging.
  • OPFS Storage Management: Intercepts, tracks, and verifies downloads. Safely purges both model files from the Origin Private File System (OPFS) when changing selections to eliminate double-storage bloat.
  • Expert-Level WebGPU Memory Management: Exposes explicit, active context cleanup APIs that interface with LiteRT's native model deallocation triggers. Instantly frees up ~2-3 GB of physical GPU textures, buffers, and WebAssembly heaps when switching models or disabling AI, preventing double-allocation spikes and Safari/Chrome tab crashes.
  • Closure-Breaking Token Stream GC: Force-nullifies stream readers and closure references (such as stream callbacks) upon completion. This aggressively triggers garbage collection in mobile JS engines to protect precious tab memory footprints.
Mobile Flashcard Mobile Tutor

8. Intelligent AI Connection Engine & Hardware Profiler

Experience top-tier, private local intelligence with advanced configuration and scheduling tools:

  • Curated High-Quality LLMs: Select from pre-configured recommended models (gemma4:e4b, llama3.2, qwen2.5, llama3, mistral, and phi3) or supply a custom model tag.
  • Model-Agnostic by Design: Gnosys-AI is no longer tied to a single primary model. The app supports shared model selection across modules, with automatic fallback to available local models when needed.
  • Hardware-Aware Recommendations: Automatically query system specifications (GPU adapter, dedicated VRAM, system RAM, Apple Silicon Unified Memory limits) to dynamically recommend optimal models.
  • VRAM Safeguards: Identifies model memory footprints in GB. Categorizes choices (✓ Recommended if fully VRAM accelerated, ⚠️ CPU Offload if offloaded to slower RAM), and disables options that exceed system thresholds to prevent crashes. One-click "Auto-Configure" automatically selects the best fit for your machine.
  • Dynamic Pull Downloader: If a chosen model is missing, Gnosys-AI automatically intercepts the state and streams the download directly within the glassmorphic loading modal using Ollama's /api/pull streaming endpoint with a real-time progress bar.
  • VRAM Pre-Loading & Connection Caching: Starts Ollama automatically (via protocol scheme fallback or Python sidecar) and preloads models into warm VRAM. Caches active connection tokens (with 4-minute invalidation checks) to eliminate chat delays.

9. Academic & Clinical Curriculum Roadmap

Expanding to support a broad array of subjects across disciplines:

  • 🧬 Natural Sciences: General Biology, Anatomy & Physiology
  • 🧮 Mathematical & Physical Sciences: Applied Physics, Wave Mechanics, SPI Physics
  • 🏥 Clinical Specialties: Ultrasound Physics & Instrumentation, Abdominal Sonography, OB/GYN Sonography, Vascular Sonography, Sectional Anatomy & Pathology
  • ✍️ Humanities & General Studies: English Composition, Ethics & Communication

🚀 Getting Started

Path A: For Students (1-Click Setup)

You do not need to download this repository or know how to code to use Gnosys-AI. The suite is hosted entirely online at https://RorriMaesu.github.io/Gnosys-AI/.

However, for Dr. Lex (the AI Tutor) to work, it runs completely free and privately on your own computer. To activate him:

  1. Install Ollama: Download and install the Ollama app for Windows (.exe) or macOS (.zip).
  2. Download the Smart Launcher:
  3. Run the Script: Open the downloaded .zip file and double-click the launcher for your system (Start_Dr_Lex.bat for Windows, or Start_Dr_Lex_Mac.command for Mac).

The script will configure CORS, detect your hardware profile, download a compatible model, and launch the website for you.


Path B: For Developers (Local Development)

Gnosys-AI is built with pure Vanilla JavaScript and HTML.

  1. Clone the repository:
    git clone https://github.com/RorriMaesu/Gnosys-AI.git
    cd Gnosys-AI
  2. Install dependencies and compile CSS:
    npm install
    npm run build:css
  3. Start the custom local HTTP server from the root directory (enables protocol handlers and hardware detection API):
    python server.py
  4. Open your browser and navigate to http://localhost:8000.

Deployment Note: The project is configured with a GitHub Actions workflow (deploy.yml) that automatically deploys the main branch to GitHub Pages.


🛠️ Architecture

  • Frontend: HTML5, TailwindCSS (via CDN), Vanilla JavaScript.
  • Data Persistence: Client-side localStorage (including Psychology mastery, gamification, and video completion state).
  • Psychology Crash Course Data Pipeline: psychology/fetch_curriculum.py regenerates psychology/video_data.js from playlist/watch metadata.
  • AI Integration: Direct asynchronous fetch requests to http://localhost:11434 (Ollama REST API).
  • Icons: FontAwesome 6.

☕ Support the Project

Gnosys-AI is a free, open-source educational suite built to help future healthcare heroes. If this app helped you pass your exams or master clinical language, consider buying me a coffee to support continued development and hosting!

Buy Me A Coffee


Built for Academic Excellence.

About

A premium local (mobile or PC) AI-powered universal study suite for college students across all disciplines.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors