logic-lab: Instructions file for Codex

Agent Instructions for Logic Lab

Commit Policy

Do not commit changes to git. The user handles all commits manually.

Agents may edit files, create directories, and run checks, but must not run git add, git commit, or other git write operations.

Implementation Preferences

• Implement straightforward tasks directly.
• Keep changes scoped to the requested domain or tool.
• Run syntax checks for changed Python files.
• Run focused tests or smoke checks when practical.
• Leave the working tree ready for user review.

Project Structure

Logic Lab contains Python and py5 translations and experiments for creative coding algorithms.

Use domain-based organization:

• physics/ for motion, forces, particles, waves, oscillation, and physical systems.
• steering_behaviors/ for autonomous agents, path following, flow fields, sensors, and swarms.
• genetic_algorithms/ for DNA, mutation, selection, and evolutionary search.
• neuro_evolution/ for neural networks evolved by genetic algorithms.
• fractals/ for recursion, L-systems, trees, Koch curves, and spatial subdivision.
• cellular_automata/ for CA grids, lattice rules, Pascal patterns, and emergence.
• mathematical/ for mathematical geometry and numerical generative systems.
• tiling_patterns/ for symmetry, tiling, textile, and deformation systems.
• research/ for experimental or hybrid systems that do not fit cleanly elsewhere.
• shared/ for reusable libraries used by multiple sketches.

Each simulation should usually follow:

domain/simulation_name/
├── simulation_name.py
├── README.md
└── screenshots/

MCP And Manifest

Logic Lab exposes selected algorithm knowledge through a local read-only MCP server in mcp/logic_lab_server.py.

The search index lives at .agents/art_manifest.json. When adding a new algorithm or simulation, agents should:

1. Add or update the simulation code and README.md.

2. Run:

   uv run python .agents/update_art_manifest.py --write