Rewrite Drawer

Graph-rewriting simulation rendered as plotter-ready linework. A test bed for autocritic — the critic-driven image evaluation system.

What It Does

• Simulates a graph rewrite rule inspired by Wolfram Physics: {{x,y},{x,z}} -> {{x,y},{x,w},{y,w},{z,w}}
• Animates growth in a browser UI with adjustable seed, frame count, event rate, layout, and random seed
• Exports the current frame as raw SVG, optimized SVG, optional vpype/vpype-gcode output, and machine-oriented G-code for a GRBL plotter server
• Can preview and send G-code over a GRBL plotter server's websocket protocol

Run It

pip install -r requirements.txt
python3 run_local.py

Then open http://127.0.0.1:8010.

Use with autocritic

Rewrite Drawer serves as a generator for autocritic's improvement loop. With both running:

# In one terminal:
python3 run_local.py

# In another:
python3 -m autocritic run --critic wolfflin --generator rewriter --model openai:gpt-5.4-mini --iterations 5

The critic evaluates each generated image through an art-theory lens and steers the simulation parameters toward more compositionally coherent results.

Notes

• The rewrite engine uses the binary graph rule {{x,y},{x,z}} -> {{x,y},{x,w},{y,w},{z,w}}
• The UI animates by grouping many rewrite events into each visible frame
• G-code export uses vpype gwrite with the profile you select. gcodemm is the safest generic default
• The app reads gcodeServer-config.txt for machine defaults if present
• Machine-oriented output assumes 1 canvas unit = 1 mm
• Exported files are written to exports/

License

MIT