Skip to content

AGENTS.md

Latest commit

 

History

History
42 lines (31 loc) · 2 KB

AGENTS.md

File metadata and controls

42 lines (31 loc) · 2 KB
Raw

Project Information for Agents

This project uses Manim Community to create mathematical animations.

Build and Configuration

Environment Setup

  • The project uses a Python virtual environment located in .venv/.
  • Manim version: v0.20.1 (Community Edition).

Testing

Since this is an animation project, "testing" primarily involves rendering scenes to verify visual correctness.

Given that that's fundamentally a visual task, a human should be the one performing validation. Thus, the AI agent should prompt the user to review and confirm whether the results are correct instead of manually testing by itself, unless stated otherwise by the human.

Code Style and Patterns

Scene Organization

All scenes must be broken into logical chunks of code within the construct method, using header comments to clearly separate each section of code, as shown below:

# --- Header 1 (high-level section)
# -- Header 2 (sub-section)
# Inline comments (short explanations for the next line or a small block)

Vector Construction Convention

When using coordinates or positions in Manim, do not construct vectors directly using np.array([x, y, z]). Instead, using directional vectors (UP, DOWN, LEFT, RIGHT, IN, OUT) scaled by a coefficient placed on the left side of the expression. For example, (3, -1) would be written as 3 * UP + DOWN.

Other Style Guidelines

  • Favor Factory Methods: The project often uses internal factory methods for consistent styling of labels and TeX elements (e.g., _label_factory, _math_tex_factory in TrigonometricRadios.py). This way, we avoid repeating initialization code over and over again.
  • Write decimal numbers without a leading zero for values between 0 and 1 (e.g., .5 instead of 0.5).
  • Avoid redundant parameter assignments (e.g., color=WHITE or run_time=1 when these are the default values). The only exception to this rule is the time in waiting commands: useself.wait(1), not self.wait().