AI tutor chat
uses active tab's web page as context
Settings
Translation
Summary
AI quiz generation from file upload, copy and paste, auto-load from webpage
ai quiz marking

🌍 Project Story: Modalis – Offline AI Learning Assistant

💭 Inspiration

Growing up in Ghana, West Africa, I witnessed a challenge that many students silently endure — the lack of reliable internet access.
In rural areas, learners are left behind, not because they lack potential, but because most modern learning tools require a constant online connection and consume large amounts of data.

High data costs and network instability often make it impossible for students to translate web content, summarize lessons, or access tutoring assistance.
That reality inspired Modalis, a Chrome extension that turns Chrome’s built-in Gemini Nano AI into a fully offline learning companion — capable of translating, summarizing, tutoring, and even generating quizzes without an internet connection.

The idea is simple but powerful:
If AI = Knowledge, then Access × Connectivity should not define Learning.
In formula form:
$$ \text{Education} = \text{AI Power} \times \text{Accessibility}^{2} $$

🧠 What I Built

Modalis is a Chrome Extension that uses Gemini Nano and Gemma3 (via Ollama) to bring offline, AI-powered learning directly into Chrome.

Key features include:

🧩 Offline Translation & Summarization powered by Chrome’s Prompt and Translator APIs
🎓 Context-Aware AI Tutor that reads and understands the webpage content (up to 8,000 characters)
📚 Interactive Quiz Generator that creates multiple-choice questions from any text or file (PDF, DOCX, PPTX, TXT)
🎤 Voice Input support for natural, hands-free learning
☁️ Gemini API Fallback ensuring uninterrupted service when local AI is unavailable

All features operate locally, protecting user privacy and eliminating dependency on internet availability.

🛠️ How I Built It

I developed Modalis using TypeScript + React within a Chrome Extension (Manifest V3) structure.

Core integrations include:

Chrome Prompt API → for Gemini Nano interactions
Chrome Translator API → for local multilingual translation
Ollama → running Gemma3 (4B) for local multimodal Q&A
Gemini API → cloud fallback
SpeechRecognition API → for real-time dictation and waveform visualization

Each module — Translator, Summarizer, Tutor, and Quiz Generator — is rendered in a unified dark UI with smooth animations and interactive elements.

⚡ What I Learned

How to work with Chrome’s experimental AI APIs, such as the Prompt and Translator APIs.
The importance of Edge AI in solving real-world educational inequalities.
How to design systems that fail gracefully, using intelligent fallbacks.
Deep insights into user privacy, AI latency optimization, and memory-efficient inference.

🚧 Challenges Faced

Sparse documentation on Chrome’s built-in AI models, requiring extensive experimentation.
Configuring Ollama with Chrome Extensions, since browser contexts restrict localhost calls.
Mic access issues, where Chrome permissions sometimes blocked voice features.
Testing in low-connectivity environments, revealing the necessity of fallback mechanisms.

Despite these hurdles, every obstacle refined the project’s purpose — to make AI learning accessible, even in the toughest conditions.

🌱 Reflection

Modalis is more than a Chrome extension; it is a statement.
From the quiet classrooms of Ghana to the broader world, it embodies one core belief:

Learning should not depend on connectivity.

With Modalis, students can translate, summarize, and learn anytime, anywhere — proving that even when the internet fades, knowledge continues to flow.

$$ \text{If\ Internet} \rightarrow 0,\ \text{then\ Learning}\ \neq\ 0 $$