"""
evaluation/
-----------
Benchmark harness and metrics for AutoDevAgent.

Standalone package — calls pipeline/graph.py directly with fixed
benchmark tasks. Reports to W&B for reproducible experiment tracking.

Modules:
    benchmark — 5 built-in tasks, BenchmarkRunner, BenchmarkResult
    metrics   — MetricsSummary, compute_metrics, format helpers

Usage:
    from evaluation.benchmark import BenchmarkRunner
    from evaluation.metrics   import compute_metrics

    runner  = BenchmarkRunner()
    results = runner.run_all()
    summary = compute_metrics(results)
    print(summary.to_summary_string())
"""

from evaluation.benchmark import (
    BenchmarkTask,
    BenchmarkResult,
    BenchmarkRunner,
    BENCHMARK_TASKS,
)
from evaluation.metrics import (
    MetricsSummary,
    compute_metrics,
    format_results_for_display,
)

__all__ = [
    "BenchmarkTask",
    "BenchmarkResult",
    "BenchmarkRunner",
    "BENCHMARK_TASKS",
    "MetricsSummary",
    "compute_metrics",
    "format_results_for_display",
]