Fractal-Down

Fractal-Down is a Python package for evaluating computational DAGs with square-root memory complexity and fractal priority scheduling. It enables running large dependency graphs on memory-constrained devices by using √N scratch memory instead of N, prioritizing high-value computation paths, and caching execution plans for deterministic replay. The library supports various use cases including on-device AI, search pipelines, code intelligence, and scientific computing, making it valuable for edge computing and resource-limited environments.https://github.com/nwoolr20/Fractal-Down