Nodus

A browser extension that transforms long-form articles into interactive knowledge graphs for faster comprehension and navigation.

GitHub Live Demo

Screens

Knowledge graph generated from an article. — Articles are transformed into nodes and relationships for exploration.

Extension side panel interface. — The interface allows users to generate and interact with the graph in real time.

Overview

Nodus reframed static articles as structured knowledge maps, allowing users to explore ideas, relationships, and summaries without linear reading.

Problem

Long-form content is dense and linear, making it difficult to quickly extract structure, identify key ideas, or revisit important sections.

Approach

I integrated a pipeline that extracts readable article content and sends it to the Gemini API for structured concept generation.

The system converts text into nodes and edges, representing relationships between ideas in an interactive graph.

The UI runs in a browser side panel, enabling real-time interaction without disrupting the reading experience.

Learnings

Structuring unstructured text is the core challenge in making AI outputs useful.
Interactive visualizations can reduce cognitive load when navigating complex information.
Running AI workflows in the browser requires careful handling of latency and user experience.

Notes

Outcome

The extension enabled users to quickly understand the structure of articles and revisit key ideas without re-reading entire pages. It demonstrated a practical use of LLMs for real-time knowledge extraction within a browser environment.

What carried forward

This project highlighted the importance of turning raw AI output into structured, navigable systems. It reinforced that usefulness comes from how information is shaped and presented, not just generated.