cosmosgl/graph

What it solves

It addresses the performance bottlenecks associated with rendering and simulating large-scale network graphs in the browser. Traditional methods often struggle with memory operations and CPU limits when handling hundreds of thousands of points and links, leading to laggy or frozen interfaces.

How it works

The engine uses WebGL (via luma.gl) to move all computations and drawing to the GPU. By utilizing fragment and vertex shaders for the force-directed graph algorithm and rendering, it avoids expensive memory transfers between the CPU and GPU, enabling real-time simulation and smooth interaction.

Who it’s for

Developers building data visualization tools, researchers analyzing complex networks, and those creating interfaces for exploring AI embeddings.

Highlights

• GPU-Accelerated Simulation: Handles hundreds of thousands of points and links in real-time.
• Collision Force: Includes a GPU-accelerated collision force using a spatial-hash grid to prevent point overlap.
• Interactive Controls: Supports dragging points, pinch-to-zoom, and comprehensive context menus for desktop and touch devices.
• Flexible Rendering: Offers control over link blending, curved links, and GPU-animated transitions for point and link attributes.
• Efficient Data Handling: Uses Float32Array for direct control over positions, colors, and sizes.