FluxRender - Python CFD & Vector Field Visualization

A high-performance engine for mathematical vector field visualization and fluid dynamics.

FluxRender is a Taichi-powered evaluation engine designed for physicists, mathematicians, and engineers. It bridges the gap between complex mathematical definitions and real-time visual analysis.

By combining a zero-redundancy math engine with a built-in Lattice Boltzmann (LBM) fluid solver, native HSL color interpolation, and adaptive spatial grids, FluxRender allows you to explore chaotic attractors, aerodynamic flows, and topological tensors interactively.

Fluid Dynamics (Physics Engine)

• Lattice Boltzmann Solver: An integrated 2D fluid dynamics solver (D2Q9) that simulates aerodynamic flows, vortices, and fluid interactions directly on the GPU.
• Solid Colliders: Embed physical obstacles into the grid using mathematical inequalities (EquationCollider) or external alpha-channel image files (ImageCollider).
• Boundary Control: Configure perimeter behaviors, including velocity inflows, open outflows, and periodic (wrapping) spaces.
• State Baking: Save and load microscopic fluid distributions as binary .npy files to bypass lengthy spin-up calculations and resume simulations instantly.

https://github.com/user-attachments/assets/6593610c-cd92-4071-ac0f-34c36b57952d

Mathematical Evaluation

• Zero-Redundancy Execution: Evaluates primary vector fields once per frame. Topological metrics (like divergence or curl) reuse pre-calculated vector data, preventing redundant GPU operations.
• Automatic Vectorization & Time Injection: Write mathematical logic in Python or NumPy. The engine inspects function signatures, handles vectorization fallbacks, and automatically injects simulation time (t).
• Interactive Data Probes: Map screen coordinates to mathematical space using CursorRegion to sample local properties or emit particles in real-time.

https://github.com/user-attachments/assets/53c9c95d-02c4-493f-ab34-731bcfe57304

Core Architecture & Visualization

• Particle Dynamics: Simulate, render, and track tens of thousands of particles driven natively by either the LBM fluid solver or custom mathematical vector fields.
• Advanced Rendering Modes: Arrow-based vector fields feature multiple rendering strategies (e.g., SCREEN_FIXED, ZOOM_DENSITY_ADAPTIVE), dynamically recalculating grid spacing based on camera zoom.
• Visual Granularity: Control the thickness, opacity, geometry, and anti-aliasing of visual elements. The ColorMapper maps scalars directly through HSL space, avoiding RGB mid-tones.
• UI Integration: A built-in UI system allows for attaching custom buttons and interaction listeners directly to the scene workspace.

https://github.com/user-attachments/assets/60d9f15f-900a-443e-af2b-55ffeb8ca285

Quick Start

See the engine in action. This minimal setup creates a fully interactive, swirling vortex, evaluated and colored dynamically based on its rotational velocity.

import FluxRender as fr
import numpy as np

# Define flow mathematics (vector function)
def flow_vector(x, y):
    X = np.sin(x) * y
    Y = np.cos(y) * x
    return X, Y

fr.quick_simulate(flow_vector) # This single line sets up a full interactive simulation with default settings.

Documentation

Dive deeper into the architecture, explore the rendering modes, and learn how to build complex physical environments in the official documentation:

👉 Read the full API Reference & Guides here