src/frontier_cs/__init__.py

"""
Frontier-CS: Evaluation framework for frontier CS problems.

Usage:
    from frontier_cs import FrontierCSEvaluator

    evaluator = FrontierCSEvaluator()

    # Algorithmic problems
    score = evaluator.evaluate("algorithmic", problem_id=1, code=cpp_code)

    # Research problems (local Docker)
    score = evaluator.evaluate("research", problem_id="flash_attn", code=py_code)

    # Research problems (SkyPilot cloud)
    score = evaluator.evaluate("research", problem_id="flash_attn", code=py_code,
                               backend="skypilot")

    # Batch evaluation with incremental progress
    from frontier_cs.batch import BatchEvaluator

    batch = BatchEvaluator(results_dir="results/gpt5")
    batch.evaluate_model("gpt-5", problems=["flash_attn", "cross_entropy"])

    # Batch evaluation with bucket storage (for SkyPilot)
    batch = BatchEvaluator(
        results_dir="results/gpt5",
        backend="skypilot",
        bucket_url="s3://my-bucket/frontier-results",
    )
    batch.# Use batch.scan_solutions_dir() or evaluate_pairs()
"""

from .evaluator import FrontierCSEvaluator
from .config import RuntimeConfig, ResourcesConfig, DockerConfig, ProblemConfig
from .runner import EvaluationResult

__all__ = [
    "FrontierCSEvaluator",
    "RuntimeConfig",
    "ResourcesConfig",
    "DockerConfig",
    "ProblemConfig",
    "EvaluationResult",
]

__version__ = "0.1.0"