Examples¶

ShinkaEvolve ships with runnable tasks demonstrating different languages, evaluation styles, and runtime profiles.

Circle Packing¶

Recommended first example.


Path	`examples/circle_packing`
Language	Python
Focus	Async evolution, config profiles, result notebooks

Key files: initial.py, evaluate.py, run_evo.py, shinka_small.yaml, shinka_medium.yaml, shinka_long.yaml

cd examples/circle_packing
python run_evo.py --config_path shinka_small.yaml

Best reference for budgeted async runs and notebook inspection after evolution.


Path	`examples/game_2048`
Language	Python
Focus	Policy optimization in a game environment

Use this for a nontrivial evaluator with task-specific environment logic and a control-oriented problem shape.


Path	`examples/julia_prime_counting`
Language	Julia candidate + Python evaluator
Focus	Cross-language evolution, strict correctness scoring

cd examples/julia_prime_counting
python evaluate.py --program_path initial.jl --results_dir results/manual_eval

Cleanest example of evolving a non-Python candidate while keeping the evaluation harness in Python.


Path	`examples/novelty_generator`
Language	Python
Focus	Novelty-oriented generation, nontraditional evaluation

Use this to inspect prompt design, novelty judgment, and nontraditional evaluation metrics.

examples/shinka_tutorial.ipynb — exploratory, notebook-centered introduction before moving into full CLI or API workflows.