from __future__ import annotations

import argparse
import json
from pathlib import Path


def _engine_from_args(args: argparse.Namespace, validation: bool = False) -> RDockEngine:
    from .rdock import RDockEngine, RDockRunConfig

    return RDockEngine(
        RDockRunConfig(
            n_runs=int(getattr(args, "n_runs", 100 if validation else 50)),
            jobs=getattr(args, "jobs", "auto"),
            cpu_fraction=float(getattr(args, "cpu_fraction", 0.85)),
            dock_prm=getattr(args, "dock_prm", None),
            rbt_root=getattr(args, "rbt_root", None),
            require_reporting_tools=validation,
        )
    )


def main() -> int:
    parser = argparse.ArgumentParser(prog="python -m docking_pipeline", description="Strict rDock-first docking pipeline")
    sub = parser.add_subparsers(dest="cmd", required=True)

    p = sub.add_parser("prepare-target")
    p.add_argument("--receptor", required=True)
    p.add_argument("--ref-ligand", required=True)
    p.add_argument("--out", required=True)
    p.add_argument("--rbt-root")
    p.add_argument("--dock-prm")

    p = sub.add_parser("dock-rdock")
    p.add_argument("--target-config", required=True)
    p.add_argument("--ligands", required=True)
    p.add_argument("--out", required=True)
    p.add_argument("--n-runs", type=int, default=50)
    p.add_argument("--jobs", default="auto")
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--dock-prm")
    p.add_argument("--rbt-root")
    p.add_argument("--dry-run", action="store_true")
    p.add_argument("--plan-only", action="store_true")

    p = sub.add_parser("validate-rdock")
    p.add_argument("--set", choices=["astex", "dud", "rna"], required=True)
    p.add_argument("--system")
    p.add_argument("--system-list")
    p.add_argument("--max-systems", type=int)
    p.add_argument("--list-systems", action="store_true")
    p.add_argument("--data-dir", required=True)
    p.add_argument("--out", default="results/rdock_validation")
    p.add_argument("--n-runs", type=int, default=100)
    p.add_argument("--jobs", default="auto")
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--download-url")
    p.add_argument("--download-if-missing", action="store_true")
    p.add_argument("--keep-work", action="store_true")
    p.add_argument("--force", action="store_true")
    p.add_argument("--dry-run", action="store_true")
    p.add_argument("--plan-only", action="store_true")

    p = sub.add_parser("run-adaptive")
    p.add_argument("--target-config", required=True)
    p.add_argument("--ligands", required=True, help="CSV with ligand_id,smiles and optional model_score")
    p.add_argument("--out", required=True)
    p.add_argument("--budget", type=int, required=True)
    p.add_argument("--batch-size", type=int, default=16)
    p.add_argument("--n-runs", type=int, default=50)
    p.add_argument("--jobs", default="auto")
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--dock-prm")
    p.add_argument("--rbt-root")
    p.add_argument("--dry-run", action="store_true")
    p.add_argument("--plan-only", action="store_true")

    p = sub.add_parser("benchmark-adaptive")
    p.add_argument("--dataset-dir", required=True)
    p.add_argument("--strategy", default="multifidelity_adaptive_rdock")
    p.add_argument("--fidelity-levels", default="5,10,15,30,50")
    p.add_argument("--cost-budget-runs", type=int, required=True)
    p.add_argument("--adaptive-budget-ligands", type=int)
    p.add_argument("--promotion-fraction", type=float, default=0.5)
    p.add_argument("--min-per-cluster", type=int, default=1)
    p.add_argument("--max-per-cluster", type=int, default=50)
    p.add_argument("--outlier-intra-z-threshold", type=float, default=3.0)
    p.add_argument("--score-component-filter", default="warn")
    p.add_argument("--final-fidelity-only-hits", default="true")
    p.add_argument("--reference-mode", default="full")
    p.add_argument("--evaluation-pool-mode", default="same_pool")
    p.add_argument("--reference-sample-size", type=int, default=5000)
    p.add_argument("--reference-sample-seed", type=int, default=42)
    p.add_argument("--balanced-baselines", default="true")
    p.add_argument("--posthoc-score-selected-hits", default="false")
    p.add_argument("--posthoc-final-runs", type=int, default=50)
    p.add_argument("--force-resume-stale", action="store_true")
    p.add_argument("--outlier-policy", default="downrank")
    p.add_argument("--intra-z-threshold", type=float, default=4.0)
    p.add_argument("--score-z-threshold", type=float, default=5.0)
    p.add_argument("--max-intra-fraction", type=float, default=0.75)
    p.add_argument("--max-intra-fraction-soft", type=float, default=0.75)
    p.add_argument("--max-intra-fraction-hard", type=float, default=0.9)
    p.add_argument("--exploration-fraction", type=float, default=0.35)
    p.add_argument("--diversity-weight", type=float, default=0.75)
    p.add_argument("--uncertainty-weight", type=float, default=0.35)
    p.add_argument("--outlier-risk-weight", type=float, default=2.0)
    p.add_argument("--cluster-min-coverage", type=int, default=1)
    p.add_argument("--use-reference-features", default="false")
    p.add_argument("--production-reference-free", default="false")
    p.add_argument("--calibration-size", type=int, default=0)
    p.add_argument("--calibration-fraction", type=float, default=0.2)
    p.add_argument("--min-clusters-covered", type=int, default=8)
    p.add_argument("--calibration-random-fraction", type=float, default=0.15)
    p.add_argument("--calibration-diversity-weight", type=float, default=1.0)
    p.add_argument("--fidelity-validation-size", type=int, default=50)
    p.add_argument("--fidelity-validation-policy", default="cluster_stratified")
    p.add_argument("--promotion-policy", default="conservative")
    p.add_argument("--min-final-ligands", type=int, default=20)
    p.add_argument("--min-promotion-per-level", type=int, default=8)
    p.add_argument("--promotion-fraction-by-level", default="")
    p.add_argument("--triage-retain-fraction", type=float, default=0.05)
    p.add_argument("--triage-target-recall", type=float, default=0.98)
    p.add_argument("--triage-min-survivors", type=int, default=50)
    p.add_argument("--triage-max-survivors", type=int, default=0)
    p.add_argument("--cluster-min-survivors", type=int, default=1)
    p.add_argument("--cluster-max-survivors", type=int, default=0)
    p.add_argument("--rescue-fraction", type=float, default=0.05)
    p.add_argument("--rare-cluster-rescue", type=int, default=20)
    p.add_argument("--uncertainty-rescue", type=int, default=20)
    p.add_argument("--allow-low-confidence-triage", action="store_true")
    p.add_argument("--top-good-fraction", type=float, default=0.1)
    p.add_argument("--minimum-training-ligands", type=int, default=50)
    p.add_argument("--triage-controller", default="auto_recall", choices=["auto_recall", "fixed"])
    p.add_argument("--max-retain-fraction-before-not-useful", type=float, default=0.5)
    p.add_argument("--classifier-top-percentile", type=float, default=0.1)
    p.add_argument("--triage-model", default="classifier")
    p.add_argument("--classifier-threshold-mode", default="recall_target")
    p.add_argument("--classifier-min-positives", type=int, default=10)
    p.add_argument("--classifier-holdout-fraction", type=float, default=0.25)
    p.add_argument("--classifier-fallback", default="cluster_only")
    p.add_argument("--model-fallback-if-worse", default="none")
    p.add_argument("--survivor-combination-policy", default="model_only")
    p.add_argument("--final-survivor-enumerate-variants", default="false")
    p.add_argument("--variant-stage", default="none")
    p.add_argument("--enumerate-stereoisomers", default="none")
    p.add_argument("--max-stereoisomers-per-parent", type=int, default=2)
    p.add_argument("--enumerate-tautomers", default="none")
    p.add_argument("--max-tautomers-per-parent", type=int, default=1)
    p.add_argument("--enumerate-protonation", default="none")
    p.add_argument("--ph", type=float, default=7.4)
    p.add_argument("--max-protomer-states-per-parent", type=int, default=1)
    p.add_argument("--max-conformers-per-variant", type=int, default=1)
    p.add_argument("--max-total-variants-per-parent", type=int, default=1)
    p.add_argument("--posthoc-top-parents", type=int, default=100)
    p.add_argument("--posthoc-max-total-variants-per-parent", type=int, default=20)
    p.add_argument("--variant-fairness-policy", default="cap")
    p.add_argument("--allow-no-rdkit-parent-only", action="store_true")
    p.add_argument("--chunk-size", type=int, default=50)
    p.add_argument("--resume", action="store_true")
    p.add_argument("--checkpoint-every", type=int, default=1)
    p.add_argument("--jobs", default="auto")
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--out", required=True)
    p.add_argument("--dry-run", action="store_true")
    p.add_argument("--plan-only", action="store_true")

    p = sub.add_parser("screen-reference-free")
    p.add_argument("--dataset-dir", required=True)
    p.add_argument("--strategy", default="reference_free_triage_bandit_v1")
    p.add_argument("--fidelity-levels", default="5,10,15,30,50")
    p.add_argument("--cost-budget-runs", type=int, required=True)
    p.add_argument("--triage-retain-fraction", type=float, default=0.02)
    p.add_argument("--triage-target-recall", type=float, default=0.98)
    p.add_argument("--calibration-size", type=int, default=1000)
    p.add_argument("--calibration-fraction", type=float, default=0.2)
    p.add_argument("--min-clusters-covered", type=int, default=8)
    p.add_argument("--calibration-random-fraction", type=float, default=0.15)
    p.add_argument("--calibration-diversity-weight", type=float, default=1.0)
    p.add_argument("--fidelity-validation-size", type=int, default=200)
    p.add_argument("--fidelity-validation-policy", default="cluster_stratified")
    p.add_argument("--promotion-policy", default="conservative")
    p.add_argument("--min-final-ligands", type=int, default=20)
    p.add_argument("--min-promotion-per-level", type=int, default=8)
    p.add_argument("--promotion-fraction-by-level", default="")
    p.add_argument("--cluster-min-survivors", type=int, default=1)
    p.add_argument("--cluster-max-survivors", type=int, default=0)
    p.add_argument("--rescue-fraction", type=float, default=0.05)
    p.add_argument("--rare-cluster-rescue", type=int, default=20)
    p.add_argument("--uncertainty-rescue", type=int, default=20)
    p.add_argument("--allow-low-confidence-triage", action="store_true")
    p.add_argument("--top-good-fraction", type=float, default=0.1)
    p.add_argument("--minimum-training-ligands", type=int, default=50)
    p.add_argument("--triage-controller", default="auto_recall", choices=["auto_recall", "fixed"])
    p.add_argument("--max-retain-fraction-before-not-useful", type=float, default=0.5)
    p.add_argument("--classifier-top-percentile", type=float, default=0.1)
    p.add_argument("--triage-model", default="classifier")
    p.add_argument("--classifier-threshold-mode", default="recall_target")
    p.add_argument("--classifier-min-positives", type=int, default=10)
    p.add_argument("--classifier-holdout-fraction", type=float, default=0.25)
    p.add_argument("--classifier-fallback", default="cluster_only")
    p.add_argument("--model-fallback-if-worse", default="none")
    p.add_argument("--survivor-combination-policy", default="model_only")
    p.add_argument("--final-survivor-enumerate-variants", default="false")
    p.add_argument("--variant-stage", default="none")
    p.add_argument("--enumerate-stereoisomers", default="none")
    p.add_argument("--max-stereoisomers-per-parent", type=int, default=2)
    p.add_argument("--enumerate-tautomers", default="none")
    p.add_argument("--max-tautomers-per-parent", type=int, default=1)
    p.add_argument("--enumerate-protonation", default="none")
    p.add_argument("--ph", type=float, default=7.4)
    p.add_argument("--max-protomer-states-per-parent", type=int, default=1)
    p.add_argument("--max-conformers-per-variant", type=int, default=1)
    p.add_argument("--max-total-variants-per-parent", type=int, default=1)
    p.add_argument("--posthoc-top-parents", type=int, default=100)
    p.add_argument("--posthoc-max-total-variants-per-parent", type=int, default=20)
    p.add_argument("--variant-fairness-policy", default="cap")
    p.add_argument("--allow-no-rdkit-parent-only", action="store_true")
    p.add_argument("--outlier-policy", default="downrank")
    p.add_argument("--intra-z-threshold", type=float, default=4.0)
    p.add_argument("--score-z-threshold", type=float, default=5.0)
    p.add_argument("--max-intra-fraction", type=float, default=0.75)
    p.add_argument("--max-intra-fraction-soft", type=float, default=0.75)
    p.add_argument("--max-intra-fraction-hard", type=float, default=0.9)
    p.add_argument("--exploration-fraction", type=float, default=0.35)
    p.add_argument("--diversity-weight", type=float, default=0.75)
    p.add_argument("--uncertainty-weight", type=float, default=0.35)
    p.add_argument("--outlier-risk-weight", type=float, default=2.0)
    p.add_argument("--cluster-min-coverage", type=int, default=1)
    p.add_argument("--jobs", default="auto")
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--chunk-size", type=int, default=50)
    p.add_argument("--rdock-timeout-seconds", type=int, default=3600)
    p.add_argument("--resume", action="store_true")
    p.add_argument("--force-resume-stale", action="store_true")
    p.add_argument("--checkpoint-every", type=int, default=1)
    p.add_argument("--out", required=True)
    p.add_argument("--dry-run", action="store_true")
    p.add_argument("--plan-only", action="store_true")

    p = sub.add_parser("screen-production-adaptive")
    p.add_argument("--dataset-dir", required=True)
    p.add_argument("--fidelity-levels", default="5,10,15,30,50")
    p.add_argument("--cost-budget-runs", type=int, required=True)
    p.add_argument("--calibration-size", type=int, default=12000)
    p.add_argument("--classifier-top-percentile", type=float, default=0.05)
    p.add_argument("--triage-target-recall", type=float, default=0.95)
    p.add_argument("--promotion-policy", default="conservative")
    p.add_argument("--min-final-ligands", type=int, default=200)
    p.add_argument("--min-promotion-per-level", type=int, default=8)
    p.add_argument("--triage-retain-fraction", type=float, default=0.05)
    p.add_argument("--cluster-min-survivors", type=int, default=1)
    p.add_argument("--cluster-max-survivors", type=int, default=0)
    p.add_argument("--rescue-fraction", type=float, default=0.05)
    p.add_argument("--rare-cluster-rescue", type=int, default=20)
    p.add_argument("--uncertainty-rescue", type=int, default=20)
    p.add_argument("--classifier-weight", type=float, default=1.0)
    p.add_argument("--regressor-weight", type=float, default=0.2)
    p.add_argument("--regressor-contribution-mode", default="gate")
    p.add_argument("--diversity-weight", type=float, default=0.75)
    p.add_argument("--uncertainty-weight", type=float, default=0.35)
    p.add_argument("--outlier-risk-weight", type=float, default=2.0)
    p.add_argument("--adaptive-policy", default="cluster_bandit")
    p.add_argument("--fixed-score-regressor-target", default="component_sane_affinity_like")
    p.add_argument("--regressor-model-type", default="extra_trees")
    p.add_argument("--model-validation-split", default="cluster", choices=["random", "cluster"])
    p.add_argument("--jobs", default="auto")
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--chunk-size", type=int, default=10)
    p.add_argument("--rdock-timeout-seconds", type=int, default=1800)
    p.add_argument("--resume", action="store_true")
    p.add_argument("--force-resume-stale", action="store_true")
    p.add_argument("--checkpoint-every", type=int, default=1)
    p.add_argument("--out", required=True)
    p.add_argument("--dry-run", action="store_true")
    p.add_argument("--plan-only", action="store_true")

    p = sub.add_parser("validate-fidelity")
    p.add_argument("--run-dir", required=True)

    p = sub.add_parser("benchmark-triage-repeated")
    p.add_argument("--dataset-dir", required=True)
    p.add_argument("--reference-mode", default="full", choices=["full", "sampled"])
    p.add_argument("--reference-sample-size", type=int, default=500)
    p.add_argument("--reference-sample-seed", type=int, default=42)
    p.add_argument("--strategy", default="reference_free_triage_bandit_v1")
    p.add_argument("--baselines", default="cluster_only_triage,cheap_descriptor_filter_only,diverse_random_cost_balanced")
    p.add_argument("--triage-target-recall", type=float, default=0.98)
    p.add_argument("--triage-retain-fraction", type=float, default=0.02)
    p.add_argument("--calibration-size", type=int, default=300)
    p.add_argument("--fidelity-validation-size", type=int, default=100)
    p.add_argument("--triage-model", default="classifier")
    p.add_argument("--classifier-top-percentile", type=float, default=0.05)
    p.add_argument("--classifier-threshold-mode", default="recall_target")
    p.add_argument("--classifier-min-positives", type=int, default=10)
    p.add_argument("--classifier-holdout-fraction", type=float, default=0.25)
    p.add_argument("--classifier-fallback", default="cluster_only")
    p.add_argument("--triage-controller", default="auto_recall")
    p.add_argument("--model-fallback-if-worse", default="none")
    p.add_argument("--survivor-combination-policy", default="model_only")
    p.add_argument("--seeds", default="1,2,3")
    p.add_argument("--jobs", default="auto")
    p.add_argument("--chunk-size", type=int, default=20)
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--out", required=True)
    p.add_argument("--force", action="store_true")

    p = sub.add_parser("benchmark-triage-strategies")
    p.add_argument("--dataset-dir", required=True)
    p.add_argument("--reference-mode", default="sampled", choices=["full", "sampled", "none"])
    p.add_argument("--reference-sample-size", type=int, default=300)
    p.add_argument("--reference-sample-seed", type=int, default=42)
    p.add_argument("--strategies", required=True)
    p.add_argument("--triage-target-recall", type=float, default=0.98)
    p.add_argument("--triage-retain-fraction", type=float, default=0.02)
    p.add_argument("--calibration-size", type=int, default=300)
    p.add_argument("--fidelity-validation-size", type=int, default=100)
    p.add_argument("--cost-budget-runs", type=int, default=3000)
    p.add_argument("--fidelity-levels", default="5,10,15,30,50")
    p.add_argument("--triage-model", default="classifier")
    p.add_argument("--classifier-top-percentile", type=float, default=0.05)
    p.add_argument("--classifier-threshold-mode", default="recall_target")
    p.add_argument("--classifier-min-positives", type=int, default=10)
    p.add_argument("--classifier-holdout-fraction", type=float, default=0.25)
    p.add_argument("--classifier-fallback", default="cluster_only")
    p.add_argument("--model-fallback-if-worse", default="union_with_cluster_only")
    p.add_argument("--survivor-combination-policy", default="union")
    p.add_argument("--adaptive-policy", default="hybrid_rank")
    p.add_argument("--adaptive-policies", default="")
    p.add_argument("--regressor-contribution-mode", default="linear")
    p.add_argument("--classifier-weight", type=float, default=1.0)
    p.add_argument("--regressor-weight", type=float, default=0.35)
    p.add_argument("--cluster-quality-weight", type=float, default=0.5)
    p.add_argument("--fixed-score-regressor-name", default="fixed_score_regressor_v1")
    p.add_argument("--fixed-score-regressor-target", default="component_sane_affinity_like")
    p.add_argument("--regressor-model-type", default="extra_trees")
    p.add_argument("--model-validation-split", default="cluster", choices=["random", "cluster"])
    p.add_argument("--promotion-policy", default="balanced")
    p.add_argument("--exploration-fraction", type=float, default=0.35)
    p.add_argument("--diversity-weight", type=float, default=0.75)
    p.add_argument("--uncertainty-weight", type=float, default=0.35)
    p.add_argument("--outlier-risk-weight", type=float, default=2.0)
    p.add_argument("--min-per-cluster", type=int, default=1)
    p.add_argument("--max-per-cluster", type=int, default=50)
    p.add_argument("--triage-max-survivors", type=int, default=0)
    p.add_argument("--cluster-quota", type=int, default=0)
    p.add_argument("--promotion-temperature", type=float, default=1.0)
    p.add_argument("--diagnostics-level", default="standard", choices=["minimal", "standard", "full"])
    p.add_argument("--classifier-gate-fraction", type=float, default=0.15)
    p.add_argument("--classifier-max-gate-fraction", type=float, default=0.2)
    p.add_argument("--seeds", default="1,2,3")
    p.add_argument("--jobs", default="10")
    p.add_argument("--chunk-size", type=int, default=20)
    p.add_argument("--rdock-timeout-seconds", type=int, default=14400)
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--out", required=True)
    p.add_argument("--force", action="store_true")
    p.add_argument("--resume", action="store_true")

    p = sub.add_parser("benchmark-regressor-ablation")
    p.add_argument("--dataset-dir", required=True)
    p.add_argument("--reference-mode", default="sampled", choices=["sampled", "none"])
    p.add_argument("--reference-sample-size", type=int, default=1000)
    p.add_argument("--reference-sample-seed", type=int, default=42)
    p.add_argument("--fidelity-levels", default="5,10,15,30,50")
    p.add_argument("--cost-budget-runs", type=int, default=3000)
    p.add_argument("--calibration-size", type=int, default=500)
    p.add_argument("--fidelity-validation-size", type=int, default=150)
    p.add_argument("--triage-target-recall", type=float, default=0.98)
    p.add_argument("--triage-retain-fraction", type=float, default=0.02)
    p.add_argument("--triage-model", default="classifier")
    p.add_argument("--classifier-top-percentile", type=float, default=0.05)
    p.add_argument("--classifier-threshold-mode", default="recall_target")
    p.add_argument("--classifier-min-positives", type=int, default=10)
    p.add_argument("--classifier-holdout-fraction", type=float, default=0.25)
    p.add_argument("--classifier-fallback", default="cluster_only")
    p.add_argument("--adaptive-policy", default="classifier_only")
    p.add_argument("--regressor-contribution-mode", default="gate")
    p.add_argument("--classifier-weight", type=float, default=1.0)
    p.add_argument("--regressor-weight", type=float, default=0.35)
    p.add_argument("--cluster-quality-weight", type=float, default=0.5)
    p.add_argument("--uncertainty-weight", type=float, default=0.35)
    p.add_argument("--diversity-weight", type=float, default=0.75)
    p.add_argument("--outlier-risk-weight", type=float, default=2.0)
    p.add_argument("--fixed-score-regressor-name", default="fixed_score_regressor_v1")
    p.add_argument("--fixed-score-regressor-target", default="component_sane_affinity_like")
    p.add_argument("--regressor-model-type", default="extra_trees")
    p.add_argument("--model-validation-split", default="cluster", choices=["random", "cluster"])
    p.add_argument("--diagnostics-level", default="standard", choices=["minimal", "standard", "full"])
    p.add_argument("--classifier-gate-fraction", type=float, default=0.15)
    p.add_argument("--classifier-max-gate-fraction", type=float, default=0.2)
    p.add_argument("--variants", default="")
    p.add_argument("--seeds", default="1,2,3")
    p.add_argument("--jobs", default="10")
    p.add_argument("--chunk-size", type=int, default=20)
    p.add_argument("--rdock-timeout-seconds", type=int, default=14400)
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--out", required=True)
    p.add_argument("--force", action="store_true")
    p.add_argument("--resume", action="store_true")

    p = sub.add_parser("benchmark-comparable-adaptive")
    p.add_argument("--dataset-dir", required=True)
    p.add_argument("--evaluation-universe-size", type=int, required=True)
    p.add_argument("--evaluation-universe-seed", type=int, default=42)
    p.add_argument("--reference-mode", default="full", choices=["full", "sampled", "none"])
    p.add_argument("--reference-sample-size", type=int, default=0)
    p.add_argument("--reference-sample-seed", type=int, default=42)
    p.add_argument("--strategies", required=True)
    p.add_argument("--cost-budget-runs", type=int, required=True)
    p.add_argument("--fidelity-levels", default="5,10,15,30,50")
    p.add_argument("--promotion-policy", default="quota_ladder")
    p.add_argument("--min-final-ligands", type=int, default=200)
    p.add_argument("--triage-target-recall", type=float, default=0.95)
    p.add_argument("--triage-retain-fraction", type=float, default=0.05)
    p.add_argument("--calibration-size", type=int, default=1000)
    p.add_argument("--fidelity-validation-size", type=int, default=100)
    p.add_argument("--adaptive-policy", default="cluster_bandit")
    p.add_argument("--regressor-contribution-mode", default="gate")
    p.add_argument("--classifier-weight", type=float, default=1.0)
    p.add_argument("--regressor-weight", type=float, default=0.2)
    p.add_argument("--cluster-quality-weight", type=float, default=0.4)
    p.add_argument("--diversity-weight", type=float, default=0.75)
    p.add_argument("--uncertainty-weight", type=float, default=0.2)
    p.add_argument("--outlier-risk-weight", type=float, default=2.0)
    p.add_argument("--classifier-top-percentile", type=float, default=0.05)
    p.add_argument("--classifier-threshold-mode", default="recall_target")
    p.add_argument("--classifier-min-positives", type=int, default=10)
    p.add_argument("--classifier-holdout-fraction", type=float, default=0.25)
    p.add_argument("--classifier-fallback", default="cluster_only")
    p.add_argument("--model-fallback-if-worse", default="none")
    p.add_argument("--survivor-combination-policy", default="model_only")
    p.add_argument("--fixed-score-regressor-name", default="fixed_score_regressor_v1")
    p.add_argument("--fixed-score-regressor-target", default="component_sane_affinity_like")
    p.add_argument("--regressor-model-type", default="extra_trees")
    p.add_argument("--model-validation-split", default="cluster", choices=["random", "cluster"])
    p.add_argument("--diagnostics-level", default="standard", choices=["minimal", "standard", "full"])
    p.add_argument("--classifier-gate-fraction", type=float, default=0.15)
    p.add_argument("--classifier-max-gate-fraction", type=float, default=0.2)
    p.add_argument("--seeds", default="1,2,3")
    p.add_argument("--jobs", default="auto")
    p.add_argument("--chunk-size", type=int, default=10)
    p.add_argument("--rdock-timeout-seconds", type=int, default=1800)
    p.add_argument("--cpu-fraction", type=float, default=0.85)
    p.add_argument("--out", required=True)
    p.add_argument("--resume", action="store_true")
    p.add_argument("--force", action="store_true")

    p = sub.add_parser("validate-dataset")
    p.add_argument("--dataset-dir", required=True)
    p.add_argument("--check-rdock-tools", action="store_true")

    p = sub.add_parser("audit-benchmark")
    p.add_argument("--run-dir", required=True)

    p = sub.add_parser("validate-benchmark-model")
    p.add_argument("--run-dir", required=True)

    p = sub.add_parser("audit-strategy-benchmark")
    p.add_argument("--run-dir", required=True)

    p = sub.add_parser("audit-regressor-ablation")
    p.add_argument("--run-dir", required=True)

    p = sub.add_parser("clean-run-cache")
    p.add_argument("--run-dir", required=True)

    args = parser.parse_args()
    if args.cmd == "prepare-target":
        engine = _engine_from_args(args)
        target = engine.prepare_target(args.receptor, args.ref_ligand, args.out)
        print(json.dumps({"target_config": str(Path(args.out) / "target_config.yaml"), "target": target.__dict__}, indent=2))
    elif args.cmd == "dock-rdock":
        from .rdock import load_target_config

        engine = _engine_from_args(args)
        if args.dry_run or args.plan_only:
            plan = {
                "target_config": args.target_config,
                "ligands": args.ligands,
                "out": args.out,
                "n_runs": args.n_runs,
                "jobs": args.jobs,
                "cpu_fraction": args.cpu_fraction,
                "commands": ["rbcavity was already run during prepare-target", "rbdock chunks will be launched by --jobs"],
            }
            Path(args.out).mkdir(parents=True, exist_ok=True)
            (Path(args.out) / "dock_plan.json").write_text(json.dumps(plan, indent=2), encoding="utf-8")
            print(json.dumps(plan, indent=2))
            return 0
        artifacts = engine.dock_sdf(load_target_config(args.target_config), args.ligands, args.out, n_runs=args.n_runs, jobs=args.jobs)
        print(json.dumps(artifacts.__dict__, indent=2))
    elif args.cmd == "validate-rdock":
        from .validation import discover_validation_systems, ensure_validation_data, validate_many

        if args.set == "rna":
            raise SystemExit("RNA rDock validation is not implemented yet; supported sets are astex and dud.")
        data_root = ensure_validation_data(args.data_dir, args.set, args.download_url, args.download_if_missing, force=args.force)
        if args.list_systems:
            systems = discover_validation_systems(data_root, args.set)
            print(json.dumps([s.__dict__ for s in systems], indent=2))
            return 0
        result = validate_many(
            args.set,
            data_root,
            args.out,
            system=args.system,
            system_list=args.system_list,
            max_systems=args.max_systems,
            n_runs=args.n_runs,
            jobs=args.jobs,
            cpu_fraction=args.cpu_fraction,
            download_url=args.download_url,
            download_if_missing=args.download_if_missing,
            force=args.force,
            dry_run=args.dry_run,
            plan_only=args.plan_only,
        )
        print(json.dumps(result, indent=2))
        metrics = result.get("metrics", {}) if isinstance(result, dict) else {}
        if metrics.get("failures"):
            return 1
    elif args.cmd == "run-adaptive":
        from .adaptive import run_adaptive
        from .rdock import load_target_config

        engine = _engine_from_args(args)
        if args.dry_run or args.plan_only:
            plan = {
                "target_config": args.target_config,
                "ligands": args.ligands,
                "out": args.out,
                "budget": args.budget,
                "batch_size": args.batch_size,
                "n_runs": args.n_runs,
                "jobs": args.jobs,
            }
            Path(args.out).mkdir(parents=True, exist_ok=True)
            (Path(args.out) / "adaptive_plan.json").write_text(json.dumps(plan, indent=2), encoding="utf-8")
            print(json.dumps(plan, indent=2))
            return 0
        run_dir = run_adaptive(load_target_config(args.target_config), args.ligands, args.out, args.budget, args.batch_size, engine, n_runs=args.n_runs)
        print(json.dumps({"run_dir": str(run_dir)}, indent=2))
    elif args.cmd == "benchmark-adaptive":
        from .benchmark_adaptive import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "screen-reference-free":
        from .benchmark_adaptive import run_reference_free_from_args

        print(json.dumps(run_reference_free_from_args(args), indent=2))
    elif args.cmd == "screen-production-adaptive":
        from .benchmark_adaptive import run_production_from_args

        print(json.dumps(run_production_from_args(args), indent=2))
    elif args.cmd == "validate-dataset":
        from .dataset import validate_dataset_dir

        print(json.dumps(validate_dataset_dir(args.dataset_dir, check_rdock_tools=args.check_rdock_tools), indent=2))
    elif args.cmd == "audit-benchmark":
        from .audit_benchmark import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "validate-fidelity":
        from .validate_fidelity import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "benchmark-triage-repeated":
        from .benchmark_triage_repeated import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "benchmark-triage-strategies":
        from .benchmark_triage_strategies import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "benchmark-regressor-ablation":
        from .benchmark_regressor_ablation import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "benchmark-comparable-adaptive":
        from .benchmark_comparable_adaptive import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "validate-benchmark-model":
        from .validate_benchmark_model import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "audit-strategy-benchmark":
        from .audit_strategy_benchmark import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "audit-regressor-ablation":
        from .audit_regressor_ablation import run_from_args

        print(json.dumps(run_from_args(args), indent=2))
    elif args.cmd == "clean-run-cache":
        from .benchmark_adaptive import clean_run_cache

        print(json.dumps(clean_run_cache(args.run_dir), indent=2))
    return 0


if __name__ == "__main__":
    try:
        raise SystemExit(main())
    except Exception as exc:
        from .provenance import RDockPipelineError

        if isinstance(exc, RDockPipelineError):
            raise SystemExit(str(exc))
        raise