@tangle-network/agent-eval 0.144.11 → 0.144.13

package/dist/experiment-tracker-CnRICnMl.js

@@ -1,500 +0,0 @@
-import { c as ValidationError } from "./errors-D-LKuDhb.js";
-import { C as mulberry32 } from "./statistics-ByxzSiOM.js";
-import { n as pairArms } from "./paired-arms-iZ08VFMN.js";
-import { n as iqr } from "./baseline-C-GocmIW.js";
-import { execSync } from "node:child_process";
-//#region src/clustered-paired-binary.ts
-/**
-* Paired binary comparison for work items nested inside independent clusters.
-*
-* Pairing is delegated to {@link pairArms}; this module adds the cluster-aware
-* estimands and inference that task-level McNemar/bootstrap utilities cannot
-* provide. Callers keep their own row shape through accessors, and every
-* matched or unpaired result returns the original row object unchanged.
-*/
-const DEFAULT_BOOTSTRAP_RESAMPLES = 1e4;
-const DEFAULT_SIGN_FLIP_RESAMPLES = 1e5;
-const MAX_RESAMPLES = 1e6;
-const DEFAULT_EXACT_CLUSTER_LIMIT = 20;
-const SIGN_FLIP_SEED_SALT = 2654435769;
-/**
-* Compare binary outcomes on matched work items while respecting independent
-* clusters. The confidence interval resamples whole clusters and recomputes the
-* task-weighted risk difference. The sign-flip test flips whole-cluster outcome
-* totals and tests that same task-weighted estimand.
-*/
-function clusteredPairedBinary(rows, options) {
-	const config = validateOptions(options);
-	const paired = pairArms(projectSelectedRows(rows, options), {
-		baselineArm: options.baselineArm,
-		treatmentArm: options.treatmentArm
-	});
-	const matchedPairs = paired.pairs.map((pair) => {
-		const baseline = pair.baseline;
-		const treatment = pair.treatment;
-		if (baseline.clusterKey !== treatment.clusterKey) throw new ValidationError(`clusteredPairedBinary: pairKey '${pair.pairKey}' rep ${pair.repIndex} crosses clusters ('${baseline.clusterKey}' vs '${treatment.clusterKey}')`);
-		return {
-			pairKey: pair.pairKey,
-			repIndex: pair.repIndex,
-			clusterKey: baseline.clusterKey,
-			baseline: baseline.original,
-			treatment: treatment.original,
-			baselinePass: baseline.pass,
-			treatmentPass: treatment.pass
-		};
-	});
-	const unpairedBaseline = paired.unpairedBaseline.map((row) => row.original);
-	const unpairedTreatment = paired.unpairedTreatment.map((row) => row.original);
-	if (matchedPairs.length === 0) return {
-		matchedPairs,
-		unpairedBaseline,
-		unpairedTreatment,
-		statistics: null
-	};
-	const clusters = summarizeClusters(matchedPairs);
-	const b10 = clusters.reduce((sum, cluster) => sum + cluster.b10, 0);
-	const b01 = clusters.reduce((sum, cluster) => sum + cluster.b01, 0);
-	const taskWeightedRiskDifference = (b10 - b01) / matchedPairs.length;
-	const equalClusterMean = mean(clusters.map((cluster) => cluster.meanDifference));
-	const bootstrap = clusters.length < 2 ? null : clusterBootstrap(clusters, config);
-	const signFlip = clusterSignFlip(clusters, config);
-	return {
-		matchedPairs,
-		unpairedBaseline,
-		unpairedTreatment,
-		statistics: {
-			nPairs: matchedPairs.length,
-			nClusters: clusters.length,
-			b10,
-			b01,
-			taskWeightedRiskDifference,
-			equalClusterMean,
-			clusters,
-			bootstrap,
-			signFlip
-		}
-	};
-}
-function validateOptions(options) {
-	assertNonEmptyString("baselineArm", options.baselineArm);
-	assertNonEmptyString("treatmentArm", options.treatmentArm);
-	if (options.baselineArm === options.treatmentArm) throw new ValidationError(`clusteredPairedBinary: baselineArm and treatmentArm are both '${options.baselineArm}'`);
-	if (!Number.isInteger(options.seed)) throw new ValidationError(`clusteredPairedBinary: seed must be an integer, got ${options.seed}`);
-	const confidence = options.confidence ?? .95;
-	if (!Number.isFinite(confidence) || confidence <= 0 || confidence >= 1) throw new ValidationError(`clusteredPairedBinary: confidence must be in (0,1), got ${confidence}`);
-	const bootstrapResamples = options.bootstrapResamples ?? DEFAULT_BOOTSTRAP_RESAMPLES;
-	assertResampleCount("bootstrapResamples", bootstrapResamples);
-	const rawMinimumBootstrapResamples = 2 / (1 - confidence);
-	const minimumBootstrapResamples = Math.ceil(rawMinimumBootstrapResamples - Number.EPSILON * Math.max(1, rawMinimumBootstrapResamples) * 8);
-	if (bootstrapResamples < minimumBootstrapResamples) throw new ValidationError(`clusteredPairedBinary: bootstrapResamples must be at least ${minimumBootstrapResamples} for confidence ${confidence} so both interval tails are represented, got ${bootstrapResamples}`);
-	const signFlipResamples = options.signFlipResamples ?? DEFAULT_SIGN_FLIP_RESAMPLES;
-	assertResampleCount("signFlipResamples", signFlipResamples);
-	const exactClusterLimit = options.exactClusterLimit ?? DEFAULT_EXACT_CLUSTER_LIMIT;
-	if (!Number.isInteger(exactClusterLimit) || exactClusterLimit < 0 || exactClusterLimit > DEFAULT_EXACT_CLUSTER_LIMIT) throw new ValidationError(`clusteredPairedBinary: exactClusterLimit must be an integer in [0,${DEFAULT_EXACT_CLUSTER_LIMIT}], got ${exactClusterLimit}`);
-	const alternative = options.alternative ?? "two-sided";
-	if (alternative !== "two-sided" && alternative !== "greater" && alternative !== "less") throw new ValidationError(`clusteredPairedBinary: alternative must be 'two-sided', 'greater', or 'less', got ${String(alternative)}`);
-	return {
-		seed: options.seed,
-		confidence,
-		bootstrapResamples,
-		alternative,
-		exactClusterLimit,
-		signFlipResamples
-	};
-}
-function assertResampleCount(name, value) {
-	if (!Number.isInteger(value) || value <= 0) throw new ValidationError(`clusteredPairedBinary: ${name} must be a positive integer, got ${value}`);
-	if (value > MAX_RESAMPLES) throw new ValidationError(`clusteredPairedBinary: ${name} must not exceed ${MAX_RESAMPLES}, got ${value}`);
-}
-function projectSelectedRows(rows, options) {
-	const projected = [];
-	for (const original of rows) {
-		const arm = options.arm(original);
-		assertNonEmptyString("arm", arm);
-		if (arm !== options.baselineArm && arm !== options.treatmentArm) continue;
-		const pairKey = options.pairKey(original);
-		const clusterKey = options.clusterKey(original);
-		const pass = options.pass(original);
-		const repKey = options.repKey?.(original);
-		assertNonEmptyString("pairKey", pairKey);
-		assertNonEmptyString("clusterKey", clusterKey);
-		if (typeof pass !== "boolean") throw new ValidationError(`clusteredPairedBinary: pass accessor must return boolean for pairKey '${pairKey}'`);
-		if (repKey !== void 0) assertNonEmptyString("repKey", repKey);
-		projected.push({
-			pairKey,
-			clusterKey,
-			arm,
-			pass,
-			repKey,
-			original
-		});
-	}
-	return projected;
-}
-function assertNonEmptyString(name, value) {
-	if (typeof value !== "string" || value.trim().length === 0) throw new ValidationError(`clusteredPairedBinary: ${name} accessor must return a non-empty string`);
-}
-function summarizeClusters(pairs) {
-	const byCluster = /* @__PURE__ */ new Map();
-	for (const pair of pairs) {
-		const summary = byCluster.get(pair.clusterKey) ?? {
-			nPairs: 0,
-			b10: 0,
-			b01: 0
-		};
-		summary.nPairs++;
-		if (pair.treatmentPass && !pair.baselinePass) summary.b10++;
-		else if (pair.baselinePass && !pair.treatmentPass) summary.b01++;
-		byCluster.set(pair.clusterKey, summary);
-	}
-	return [...byCluster.entries()].sort(([a], [b]) => a < b ? -1 : a > b ? 1 : 0).map(([clusterKey, summary]) => ({
-		clusterKey,
-		...summary,
-		meanDifference: (summary.b10 - summary.b01) / summary.nPairs
-	}));
-}
-function clusterBootstrap(clusters, config) {
-	const rng = mulberry32(config.seed);
-	const samples = new Array(config.bootstrapResamples);
-	for (let draw = 0; draw < config.bootstrapResamples; draw++) {
-		let differenceSum = 0;
-		let pairCount = 0;
-		for (let index = 0; index < clusters.length; index++) {
-			const cluster = clusters[Math.floor(rng() * clusters.length)];
-			differenceSum += cluster.b10 - cluster.b01;
-			pairCount += cluster.nPairs;
-		}
-		samples[draw] = differenceSum / pairCount;
-	}
-	samples.sort((a, b) => a - b);
-	const alpha = 1 - config.confidence;
-	const lowerIndex = Math.floor(alpha / 2 * config.bootstrapResamples);
-	const upperIndex = Math.min(config.bootstrapResamples - 1, Math.ceil((1 - alpha / 2) * config.bootstrapResamples) - 1);
-	return {
-		statistic: "task-weighted-risk-difference",
-		lower: samples[lowerIndex],
-		upper: samples[Math.max(lowerIndex, upperIndex)],
-		confidence: config.confidence,
-		resamples: config.bootstrapResamples,
-		seed: config.seed
-	};
-}
-function clusterSignFlip(clusters, config) {
-	const clusterTotals = clusters.map((cluster) => cluster.b10 - cluster.b01);
-	const nonZero = clusterTotals.filter((delta) => delta !== 0);
-	const totalPairs = clusters.reduce((sum, cluster) => sum + cluster.nPairs, 0);
-	const statistic = clusterTotals.reduce((sum, delta) => sum + delta, 0) / totalPairs;
-	if (nonZero.length <= config.exactClusterLimit) {
-		const assignments = 2 ** nonZero.length;
-		let extreme = 0;
-		for (let mask = 0; mask < assignments; mask++) {
-			let sum = 0;
-			for (let index = 0; index < nonZero.length; index++) sum += (mask & 2 ** index ? 1 : -1) * nonZero[index];
-			if (isExtreme(sum / totalPairs, statistic, config.alternative)) extreme++;
-		}
-		return {
-			statistic,
-			pValue: extreme / assignments,
-			alternative: config.alternative,
-			method: "exact",
-			assignments,
-			nClusters: clusters.length,
-			nNonZeroClusters: nonZero.length,
-			seed: null
-		};
-	}
-	const signFlipSeed = (config.seed ^ SIGN_FLIP_SEED_SALT) >>> 0;
-	const rng = mulberry32(signFlipSeed);
-	let extreme = 0;
-	for (let draw = 0; draw < config.signFlipResamples; draw++) {
-		let sum = 0;
-		for (const delta of nonZero) sum += (rng() < .5 ? -1 : 1) * delta;
-		if (isExtreme(sum / totalPairs, statistic, config.alternative)) extreme++;
-	}
-	return {
-		statistic,
-		pValue: (extreme + 1) / (config.signFlipResamples + 1),
-		alternative: config.alternative,
-		method: "monte-carlo",
-		assignments: config.signFlipResamples,
-		nClusters: clusters.length,
-		nNonZeroClusters: nonZero.length,
-		seed: signFlipSeed
-	};
-}
-function isExtreme(candidate, observed, alternative) {
-	const tolerance = Number.EPSILON * Math.max(1, Math.abs(observed)) * 16;
-	if (alternative === "greater") return candidate >= observed - tolerance;
-	if (alternative === "less") return candidate <= observed + tolerance;
-	return Math.abs(candidate) >= Math.abs(observed) - tolerance;
-}
-function mean(values) {
-	return values.reduce((sum, value) => sum + value, 0) / values.length;
-}
-//#endregion
-//#region src/experiment-tracker.ts
-/**
-* Experiment tracker — git-provenanced experiment log with N-rep stats and a
-* KEEP / REGRESSION / NOISE verdict against a parent.
-*
-* Every loop the fleet runs reduces to the same question: "I ran the candidate
-* N times — is the median measurably better than the parent, or is the delta
-* inside the noise band?" The hand-rolled copies bake a fixed score scale
-* (percentage points), a fixed store path (`.evolve/experiments-v2.json`), and
-* `execSync('git …')` straight into the module. This is the canonical version:
-* provenance and persistence are injected, thresholds are configurable, and the
-* stats + verdict are pure functions you can unit-test without a git repo or a
-* filesystem.
-*
-* Stats per experiment: median / mean / min / max / iqr / stddev / passRate /
-* n, plus a `stable` flag (`iqr < iqrUnstableAbove && stddev < stddevUnstableAbove`).
-*
-* Verdict against a parent (both must have `n >= minRepsForVerdict`):
-*   - NOISE       — the candidate is too unstable to judge (`!stable`)
-*   - KEEP        — `medianDelta >  keepThreshold`
-*   - REGRESSION  — `medianDelta < -regressionThreshold`
-*   - NOISE       — otherwise (delta inside the band)
-* With no parent (or insufficient reps) the verdict is the neutral ITERATE.
-*/
-const DEFAULTS = {
-	keepThreshold: 5,
-	regressionThreshold: 5,
-	iqrUnstableAbove: 10,
-	stddevUnstableAbove: Number.POSITIVE_INFINITY,
-	minRepsForVerdict: 3
-};
-function resolveThresholds(t) {
-	const r = {
-		...DEFAULTS,
-		...t ?? {}
-	};
-	if (r.keepThreshold < 0) throw new ValidationError(`experiment-tracker: keepThreshold must be >= 0, got ${r.keepThreshold}`);
-	if (r.regressionThreshold < 0) throw new ValidationError(`experiment-tracker: regressionThreshold must be >= 0, got ${r.regressionThreshold}`);
-	if (r.minRepsForVerdict < 1) throw new ValidationError(`experiment-tracker: minRepsForVerdict must be >= 1, got ${r.minRepsForVerdict}`);
-	return r;
-}
-function median(sorted) {
-	const n = sorted.length;
-	if (n === 0) return 0;
-	const mid = Math.floor(n / 2);
-	return n % 2 === 0 ? (sorted[mid - 1] + sorted[mid]) / 2 : sorted[mid];
-}
-/** Population standard deviation (÷n). 0 for fewer than 2 values. */
-function stddev(values, mean) {
-	if (values.length < 2) return 0;
-	const variance = values.reduce((acc, v) => acc + (v - mean) ** 2, 0) / values.length;
-	return Math.sqrt(variance);
-}
-/**
-* Compute the N-rep statistics for a set of reps. Pure — no I/O. The `stable`
-* flag is the trust gate the verdict depends on: a sample whose spread exceeds
-* the configured bounds can't distinguish a real delta from run-to-run noise.
-*/
-function computeExperimentStats(reps, thresholds) {
-	const t = resolveThresholds(thresholds);
-	const n = reps.length;
-	if (n === 0) return {
-		median: 0,
-		mean: 0,
-		min: 0,
-		max: 0,
-		iqr: 0,
-		stddev: 0,
-		passRate: null,
-		n: 0,
-		stable: false
-	};
-	const scores = reps.map((r) => {
-		if (!Number.isFinite(r.score)) throw new ValidationError(`experiment-tracker: rep ${r.rep} has non-finite score ${r.score}`);
-		return r.score;
-	});
-	const sorted = [...scores].sort((a, b) => a - b);
-	const mean = scores.reduce((s, v) => s + v, 0) / n;
-	const sd = stddev(scores, mean);
-	const spread = iqr(scores);
-	const rated = reps.filter((r) => typeof r.passed === "boolean");
-	const passRate = rated.length === 0 ? null : rated.filter((r) => r.passed).length / rated.length;
-	const stable = spread < t.iqrUnstableAbove && sd < t.stddevUnstableAbove;
-	return {
-		median: median(sorted),
-		mean,
-		min: sorted[0],
-		max: sorted[n - 1],
-		iqr: spread,
-		stddev: sd,
-		passRate,
-		n,
-		stable
-	};
-}
-/**
-* Verdict for a candidate against its parent. Pure — operates on already-computed
-* stats. KEEP/REGRESSION require both sides to have `>= minRepsForVerdict` reps
-* AND the candidate to be `stable`; otherwise the result is NOISE (unstable) or
-* ITERATE (not enough reps / no parent).
-*/
-function improvementVerdict(candidate, parent, thresholds) {
-	const t = resolveThresholds(thresholds);
-	if (!parent) return {
-		verdict: "ITERATE",
-		medianDelta: null,
-		reason: "no parent experiment to compare against"
-	};
-	if (candidate.n < t.minRepsForVerdict || parent.n < t.minRepsForVerdict) return {
-		verdict: "ITERATE",
-		medianDelta: null,
-		reason: `need >= ${t.minRepsForVerdict} reps on both sides (candidate n=${candidate.n}, parent n=${parent.n})`
-	};
-	if (!candidate.stable) return {
-		verdict: "NOISE",
-		medianDelta: candidate.median - parent.median,
-		reason: `candidate unstable (iqr=${candidate.iqr}, stddev=${candidate.stddev.toFixed(2)})`
-	};
-	const medianDelta = candidate.median - parent.median;
-	if (medianDelta > t.keepThreshold) return {
-		verdict: "KEEP",
-		medianDelta,
-		reason: `median +${medianDelta} > +${t.keepThreshold}`
-	};
-	if (medianDelta < -t.regressionThreshold) return {
-		verdict: "REGRESSION",
-		medianDelta,
-		reason: `median ${medianDelta} < -${t.regressionThreshold}`
-	};
-	return {
-		verdict: "NOISE",
-		medianDelta,
-		reason: `median delta ${medianDelta} inside noise band [-${t.regressionThreshold}, +${t.keepThreshold}]`
-	};
-}
-/**
-* Default provenance reader: `git rev-parse HEAD`, the subject line, and the
-* files changed vs `HEAD~1`. Fail-loud — a tracker that silently logs
-* `commit: 'unknown'` corrupts the provenance the whole point of the log is to
-* carry. When the working tree genuinely has no parent commit, pass an override.
-*/
-const gitProvenanceReader = () => {
-	const run = (cmd) => execSync(cmd, { encoding: "utf8" }).trim();
-	const commit = run("git rev-parse --short HEAD");
-	const message = run("git log -1 --format=%s");
-	const changedRaw = run("git diff --name-only HEAD~1");
-	return {
-		commit,
-		message,
-		changedFiles: changedRaw.length === 0 ? [] : changedRaw.split("\n").filter(Boolean)
-	};
-};
-/** In-memory store — the default when no persistence is wanted (tests, ephemeral
-*  runs). State lives on the instance. */
-function inMemoryExperimentStore(initial = []) {
-	let state = initial.map((e) => structuredClone(e));
-	return {
-		async load() {
-			return state.map((e) => structuredClone(e));
-		},
-		async save(experiments) {
-			state = experiments.map((e) => structuredClone(e));
-		}
-	};
-}
-/** Filesystem store — a single JSON array at `path`, created on first save. */
-function fileExperimentStore(path) {
-	return {
-		async load() {
-			const fs = await import("node:fs/promises");
-			try {
-				const raw = await fs.readFile(path, "utf8");
-				const parsed = JSON.parse(raw);
-				if (!Array.isArray(parsed)) throw new ValidationError(`experiment-tracker: store at ${path} is not a JSON array`);
-				return parsed;
-			} catch (err) {
-				if (err.code === "ENOENT") return [];
-				throw err;
-			}
-		},
-		async save(experiments) {
-			const fs = await import("node:fs/promises");
-			const pathMod = await import("node:path");
-			await fs.mkdir(pathMod.dirname(path), { recursive: true });
-			await fs.writeFile(path, JSON.stringify(experiments, null, 2), "utf8");
-		}
-	};
-}
-/**
-* Stateful tracker over an `ExperimentStore`. Create an experiment (provenance
-* is captured once), append reps as they complete (stats + verdict recompute on
-* every append), and read the log back for a dashboard. All persistence and git
-* access flow through the injected seams, so the tracker is fully testable
-* without a repo or disk.
-*/
-var ExperimentTracker = class {
-	store;
-	provenanceReader;
-	thresholds;
-	now;
-	constructor(options = {}) {
-		this.store = options.store ?? inMemoryExperimentStore();
-		this.provenanceReader = options.provenanceReader ?? gitProvenanceReader;
-		this.thresholds = resolveThresholds(options.thresholds);
-		this.now = options.now ?? Date.now;
-	}
-	async create(input) {
-		const experiments = await this.store.load();
-		if (experiments.some((e) => e.id === input.id)) throw new ValidationError(`experiment-tracker: experiment id "${input.id}" already exists`);
-		if (input.parentId && !experiments.some((e) => e.id === input.parentId)) throw new ValidationError(`experiment-tracker: parent experiment "${input.parentId}" not found`);
-		const provenance = input.provenance ?? await this.provenanceReader();
-		const experiment = {
-			id: input.id,
-			label: input.label,
-			provenance,
-			parentId: input.parentId,
-			changeSummary: input.changeSummary,
-			reps: [],
-			stats: computeExperimentStats([], this.thresholds),
-			verdict: "ITERATE",
-			createdAt: new Date(this.now()).toISOString()
-		};
-		experiments.push(experiment);
-		await this.store.save(experiments);
-		return structuredClone(experiment);
-	}
-	/** Append a rep (its `rep` index defaults to the current rep count) and
-	*  recompute stats + verdict. Returns the updated experiment. */
-	async addRep(experimentId, rep) {
-		const experiments = await this.store.load();
-		const exp = experiments.find((e) => e.id === experimentId);
-		if (!exp) throw new ValidationError(`experiment-tracker: experiment "${experimentId}" not found`);
-		const fullRep = {
-			rep: rep.rep ?? exp.reps.length,
-			score: rep.score,
-			passed: rep.passed,
-			metrics: rep.metrics,
-			timestamp: rep.timestamp ?? new Date(this.now()).toISOString()
-		};
-		exp.reps.push(fullRep);
-		exp.stats = computeExperimentStats(exp.reps, this.thresholds);
-		const parent = exp.parentId ? experiments.find((e) => e.id === exp.parentId) : void 0;
-		exp.verdict = improvementVerdict(exp.stats, parent?.stats ?? null, this.thresholds).verdict;
-		await this.store.save(experiments);
-		return structuredClone(exp);
-	}
-	async get(experimentId) {
-		const found = (await this.store.load()).find((e) => e.id === experimentId);
-		return found ? structuredClone(found) : void 0;
-	}
-	async list() {
-		return this.store.load();
-	}
-	/** Full verdict (not just the enum) for an experiment vs its parent. */
-	async verdictFor(experimentId) {
-		const experiments = await this.store.load();
-		const exp = experiments.find((e) => e.id === experimentId);
-		if (!exp) throw new ValidationError(`experiment-tracker: experiment "${experimentId}" not found`);
-		const parent = exp.parentId ? experiments.find((e) => e.id === exp.parentId) : void 0;
-		return improvementVerdict(exp.stats, parent?.stats ?? null, this.thresholds);
-	}
-};
-//#endregion
-export { improvementVerdict as a, gitProvenanceReader as i, computeExperimentStats as n, inMemoryExperimentStore as o, fileExperimentStore as r, clusteredPairedBinary as s, ExperimentTracker as t };
-
-//# sourceMappingURL=experiment-tracker-CnRICnMl.js.map

package/dist/experiment-tracker-CnRICnMl.js.map

@@ -1 +0,0 @@
-{"version":3,"file":"experiment-tracker-CnRICnMl.js","names":[],"sources":["../src/clustered-paired-binary.ts","../src/experiment-tracker.ts"],"sourcesContent":["/**\n * Paired binary comparison for work items nested inside independent clusters.\n *\n * Pairing is delegated to {@link pairArms}; this module adds the cluster-aware\n * estimands and inference that task-level McNemar/bootstrap utilities cannot\n * provide. Callers keep their own row shape through accessors, and every\n * matched or unpaired result returns the original row object unchanged.\n */\n\nimport { ValidationError } from './errors'\nimport { type PairedArmRow, pairArms } from './paired-arms'\nimport { mulberry32 } from './statistics'\n\nexport type ClusterSignFlipAlternative = 'two-sided' | 'greater' | 'less'\n\nexport interface ClusteredPairedBinaryOptions<TRow> {\n  /** Arm treated as the control side of every pair. */\n  baselineArm: string\n  /** Arm treated as the treatment side of every pair. */\n  treatmentArm: string\n  /** Stable work-item identity shared by both arms. */\n  pairKey: (row: TRow) => string\n  /** Independent-cluster identity, shared by both rows in a matched pair. */\n  clusterKey: (row: TRow) => string\n  /** Arm identity for this row. Rows from other arms are ignored. */\n  arm: (row: TRow) => string\n  /** Binary outcome. */\n  pass: (row: TRow) => boolean\n  /** Replicate identity when a work item has multiple rows in either arm. */\n  repKey?: (row: TRow) => string | undefined\n  /** Deterministic seed for bootstrap and Monte Carlo sign-flip draws. */\n  seed: number\n  /** Percentile confidence level. Default 0.95. */\n  confidence?: number\n  /** Whole-cluster bootstrap draws. Default 10,000; maximum 1,000,000. */\n  bootstrapResamples?: number\n  /** Sign-flip alternative. Default 'two-sided'. */\n  alternative?: ClusterSignFlipAlternative\n  /**\n   * Enumerate every sign assignment at or below this many non-zero clusters;\n   * otherwise use Monte Carlo. Default 20; maximum 20.\n   */\n  exactClusterLimit?: number\n  /** Monte Carlo sign-flip draws when exact enumeration is not used. Default 100,000; maximum 1,000,000. */\n  signFlipResamples?: number\n}\n\nexport interface ClusteredMatchedPair<TRow> {\n  pairKey: string\n  repIndex: number\n  clusterKey: string\n  baseline: TRow\n  treatment: TRow\n  baselinePass: boolean\n  treatmentPass: boolean\n}\n\nexport interface ClusteredBinaryCluster {\n  clusterKey: string\n  nPairs: number\n  /** Treatment passes and baseline fails. */\n  b10: number\n  /** Baseline passes and treatment fails. */\n  b01: number\n  /** Mean (treatment - baseline) binary outcome within this cluster. */\n  meanDifference: number\n}\n\nexport interface ClusterBootstrapInterval {\n  /** The interval resamples clusters and recomputes this task-weighted statistic. */\n  statistic: 'task-weighted-risk-difference'\n  lower: number\n  upper: number\n  confidence: number\n  resamples: number\n  seed: number\n}\n\nexport interface ClusterSignFlipResult {\n  /** Task-weighted paired risk difference, matching the reported bootstrap estimand. */\n  statistic: number\n  /**\n   * Randomization p-value under whole-cluster arm-label exchangeability.\n   * `method: 'exact'` means every cluster-level sign assignment was enumerated;\n   * it does not make the exchangeability assumption unnecessary.\n   */\n  pValue: number\n  alternative: ClusterSignFlipAlternative\n  method: 'exact' | 'monte-carlo'\n  /** Exact assignments enumerated or Monte Carlo assignments drawn. */\n  assignments: number\n  nClusters: number\n  nNonZeroClusters: number\n  /** Null for exact enumeration, which has no random draws. */\n  seed: number | null\n}\n\nexport interface ClusteredPairedBinaryStatistics {\n  nPairs: number\n  nClusters: number\n  /** Treatment passes and baseline fails, across all matched pairs. */\n  b10: number\n  /** Baseline passes and treatment fails, across all matched pairs. */\n  b01: number\n  /** Mean paired difference across tasks, so every task has equal weight. */\n  taskWeightedRiskDifference: number\n  /** Mean of cluster-level paired differences, so every cluster has equal weight. */\n  equalClusterMean: number\n  clusters: ClusteredBinaryCluster[]\n  /** Null below two independent clusters; one cluster cannot estimate cluster uncertainty. */\n  bootstrap: ClusterBootstrapInterval | null\n  signFlip: ClusterSignFlipResult\n}\n\nexport interface ClusteredPairedBinaryResult<TRow> {\n  matchedPairs: ClusteredMatchedPair<TRow>[]\n  unpairedBaseline: TRow[]\n  unpairedTreatment: TRow[]\n  /** Null when there are no matched rows; absence is never reported as a zero effect. */\n  statistics: ClusteredPairedBinaryStatistics | null\n}\n\ninterface ProjectedRow<TRow> extends PairedArmRow {\n  clusterKey: string\n  original: TRow\n}\n\nconst DEFAULT_BOOTSTRAP_RESAMPLES = 10_000\nconst DEFAULT_SIGN_FLIP_RESAMPLES = 100_000\nconst MAX_RESAMPLES = 1_000_000\nconst DEFAULT_EXACT_CLUSTER_LIMIT = 20\nconst SIGN_FLIP_SEED_SALT = 0x9e3779b9\n\n/**\n * Compare binary outcomes on matched work items while respecting independent\n * clusters. The confidence interval resamples whole clusters and recomputes the\n * task-weighted risk difference. The sign-flip test flips whole-cluster outcome\n * totals and tests that same task-weighted estimand.\n */\nexport function clusteredPairedBinary<TRow>(\n  rows: readonly TRow[],\n  options: ClusteredPairedBinaryOptions<TRow>,\n): ClusteredPairedBinaryResult<TRow> {\n  const config = validateOptions(options)\n  const projected = projectSelectedRows(rows, options)\n  const paired = pairArms(projected, {\n    baselineArm: options.baselineArm,\n    treatmentArm: options.treatmentArm,\n  })\n\n  const matchedPairs = paired.pairs.map((pair): ClusteredMatchedPair<TRow> => {\n    const baseline = pair.baseline as ProjectedRow<TRow>\n    const treatment = pair.treatment as ProjectedRow<TRow>\n    if (baseline.clusterKey !== treatment.clusterKey) {\n      throw new ValidationError(\n        `clusteredPairedBinary: pairKey '${pair.pairKey}' rep ${pair.repIndex} crosses clusters ` +\n          `('${baseline.clusterKey}' vs '${treatment.clusterKey}')`,\n      )\n    }\n    return {\n      pairKey: pair.pairKey,\n      repIndex: pair.repIndex,\n      clusterKey: baseline.clusterKey,\n      baseline: baseline.original,\n      treatment: treatment.original,\n      baselinePass: baseline.pass!,\n      treatmentPass: treatment.pass!,\n    }\n  })\n\n  const unpairedBaseline = paired.unpairedBaseline.map(\n    (row) => (row as ProjectedRow<TRow>).original,\n  )\n  const unpairedTreatment = paired.unpairedTreatment.map(\n    (row) => (row as ProjectedRow<TRow>).original,\n  )\n\n  if (matchedPairs.length === 0) {\n    return { matchedPairs, unpairedBaseline, unpairedTreatment, statistics: null }\n  }\n\n  const clusters = summarizeClusters(matchedPairs)\n  const b10 = clusters.reduce((sum, cluster) => sum + cluster.b10, 0)\n  const b01 = clusters.reduce((sum, cluster) => sum + cluster.b01, 0)\n  const taskWeightedRiskDifference = (b10 - b01) / matchedPairs.length\n  const equalClusterMean = mean(clusters.map((cluster) => cluster.meanDifference))\n  const bootstrap = clusters.length < 2 ? null : clusterBootstrap(clusters, config)\n  const signFlip = clusterSignFlip(clusters, config)\n\n  return {\n    matchedPairs,\n    unpairedBaseline,\n    unpairedTreatment,\n    statistics: {\n      nPairs: matchedPairs.length,\n      nClusters: clusters.length,\n      b10,\n      b01,\n      taskWeightedRiskDifference,\n      equalClusterMean,\n      clusters,\n      bootstrap,\n      signFlip,\n    },\n  }\n}\n\ninterface ValidatedConfig {\n  seed: number\n  confidence: number\n  bootstrapResamples: number\n  alternative: ClusterSignFlipAlternative\n  exactClusterLimit: number\n  signFlipResamples: number\n}\n\nfunction validateOptions<TRow>(options: ClusteredPairedBinaryOptions<TRow>): ValidatedConfig {\n  assertNonEmptyString('baselineArm', options.baselineArm)\n  assertNonEmptyString('treatmentArm', options.treatmentArm)\n  if (options.baselineArm === options.treatmentArm) {\n    throw new ValidationError(\n      `clusteredPairedBinary: baselineArm and treatmentArm are both '${options.baselineArm}'`,\n    )\n  }\n  if (!Number.isInteger(options.seed)) {\n    throw new ValidationError(`clusteredPairedBinary: seed must be an integer, got ${options.seed}`)\n  }\n  const confidence = options.confidence ?? 0.95\n  if (!Number.isFinite(confidence) || confidence <= 0 || confidence >= 1) {\n    throw new ValidationError(\n      `clusteredPairedBinary: confidence must be in (0,1), got ${confidence}`,\n    )\n  }\n  const bootstrapResamples = options.bootstrapResamples ?? DEFAULT_BOOTSTRAP_RESAMPLES\n  assertResampleCount('bootstrapResamples', bootstrapResamples)\n  const rawMinimumBootstrapResamples = 2 / (1 - confidence)\n  const minimumBootstrapResamples = Math.ceil(\n    rawMinimumBootstrapResamples - Number.EPSILON * Math.max(1, rawMinimumBootstrapResamples) * 8,\n  )\n  if (bootstrapResamples < minimumBootstrapResamples) {\n    throw new ValidationError(\n      `clusteredPairedBinary: bootstrapResamples must be at least ${minimumBootstrapResamples} for confidence ${confidence} so both interval tails are represented, got ${bootstrapResamples}`,\n    )\n  }\n  const signFlipResamples = options.signFlipResamples ?? DEFAULT_SIGN_FLIP_RESAMPLES\n  assertResampleCount('signFlipResamples', signFlipResamples)\n  const exactClusterLimit = options.exactClusterLimit ?? DEFAULT_EXACT_CLUSTER_LIMIT\n  if (\n    !Number.isInteger(exactClusterLimit) ||\n    exactClusterLimit < 0 ||\n    exactClusterLimit > DEFAULT_EXACT_CLUSTER_LIMIT\n  ) {\n    throw new ValidationError(\n      `clusteredPairedBinary: exactClusterLimit must be an integer in [0,${DEFAULT_EXACT_CLUSTER_LIMIT}], got ${exactClusterLimit}`,\n    )\n  }\n  const alternative = options.alternative ?? 'two-sided'\n  if (alternative !== 'two-sided' && alternative !== 'greater' && alternative !== 'less') {\n    throw new ValidationError(\n      `clusteredPairedBinary: alternative must be 'two-sided', 'greater', or 'less', got ${String(alternative)}`,\n    )\n  }\n  return {\n    seed: options.seed,\n    confidence,\n    bootstrapResamples,\n    alternative,\n    exactClusterLimit,\n    signFlipResamples,\n  }\n}\n\nfunction assertResampleCount(name: string, value: number): void {\n  if (!Number.isInteger(value) || value <= 0) {\n    throw new ValidationError(\n      `clusteredPairedBinary: ${name} must be a positive integer, got ${value}`,\n    )\n  }\n  if (value > MAX_RESAMPLES) {\n    throw new ValidationError(\n      `clusteredPairedBinary: ${name} must not exceed ${MAX_RESAMPLES}, got ${value}`,\n    )\n  }\n}\n\nfunction projectSelectedRows<TRow>(\n  rows: readonly TRow[],\n  options: ClusteredPairedBinaryOptions<TRow>,\n): ProjectedRow<TRow>[] {\n  const projected: ProjectedRow<TRow>[] = []\n  for (const original of rows) {\n    const arm = options.arm(original)\n    assertNonEmptyString('arm', arm)\n    if (arm !== options.baselineArm && arm !== options.treatmentArm) continue\n\n    const pairKey = options.pairKey(original)\n    const clusterKey = options.clusterKey(original)\n    const pass = options.pass(original)\n    const repKey = options.repKey?.(original)\n    assertNonEmptyString('pairKey', pairKey)\n    assertNonEmptyString('clusterKey', clusterKey)\n    if (typeof pass !== 'boolean') {\n      throw new ValidationError(\n        `clusteredPairedBinary: pass accessor must return boolean for pairKey '${pairKey}'`,\n      )\n    }\n    if (repKey !== undefined) assertNonEmptyString('repKey', repKey)\n    projected.push({ pairKey, clusterKey, arm, pass, repKey, original })\n  }\n  return projected\n}\n\nfunction assertNonEmptyString(name: string, value: string): void {\n  if (typeof value !== 'string' || value.trim().length === 0) {\n    throw new ValidationError(\n      `clusteredPairedBinary: ${name} accessor must return a non-empty string`,\n    )\n  }\n}\n\nfunction summarizeClusters<TRow>(\n  pairs: readonly ClusteredMatchedPair<TRow>[],\n): ClusteredBinaryCluster[] {\n  const byCluster = new Map<string, { nPairs: number; b10: number; b01: number }>()\n  for (const pair of pairs) {\n    const summary = byCluster.get(pair.clusterKey) ?? { nPairs: 0, b10: 0, b01: 0 }\n    summary.nPairs++\n    if (pair.treatmentPass && !pair.baselinePass) summary.b10++\n    else if (pair.baselinePass && !pair.treatmentPass) summary.b01++\n    byCluster.set(pair.clusterKey, summary)\n  }\n  return [...byCluster.entries()]\n    .sort(([a], [b]) => (a < b ? -1 : a > b ? 1 : 0))\n    .map(([clusterKey, summary]) => ({\n      clusterKey,\n      ...summary,\n      meanDifference: (summary.b10 - summary.b01) / summary.nPairs,\n    }))\n}\n\nfunction clusterBootstrap(\n  clusters: readonly ClusteredBinaryCluster[],\n  config: ValidatedConfig,\n): ClusterBootstrapInterval {\n  const rng = mulberry32(config.seed)\n  const samples = new Array<number>(config.bootstrapResamples)\n  for (let draw = 0; draw < config.bootstrapResamples; draw++) {\n    let differenceSum = 0\n    let pairCount = 0\n    for (let index = 0; index < clusters.length; index++) {\n      const cluster = clusters[Math.floor(rng() * clusters.length)]!\n      differenceSum += cluster.b10 - cluster.b01\n      pairCount += cluster.nPairs\n    }\n    samples[draw] = differenceSum / pairCount\n  }\n  samples.sort((a, b) => a - b)\n  const alpha = 1 - config.confidence\n  const lowerIndex = Math.floor((alpha / 2) * config.bootstrapResamples)\n  const upperIndex = Math.min(\n    config.bootstrapResamples - 1,\n    Math.ceil((1 - alpha / 2) * config.bootstrapResamples) - 1,\n  )\n  return {\n    statistic: 'task-weighted-risk-difference',\n    lower: samples[lowerIndex]!,\n    upper: samples[Math.max(lowerIndex, upperIndex)]!,\n    confidence: config.confidence,\n    resamples: config.bootstrapResamples,\n    seed: config.seed,\n  }\n}\n\nfunction clusterSignFlip(\n  clusters: readonly ClusteredBinaryCluster[],\n  config: ValidatedConfig,\n): ClusterSignFlipResult {\n  const clusterTotals = clusters.map((cluster) => cluster.b10 - cluster.b01)\n  const nonZero = clusterTotals.filter((delta) => delta !== 0)\n  const totalPairs = clusters.reduce((sum, cluster) => sum + cluster.nPairs, 0)\n  const statistic = clusterTotals.reduce((sum, delta) => sum + delta, 0) / totalPairs\n  if (nonZero.length <= config.exactClusterLimit) {\n    const assignments = 2 ** nonZero.length\n    let extreme = 0\n    for (let mask = 0; mask < assignments; mask++) {\n      let sum = 0\n      for (let index = 0; index < nonZero.length; index++) {\n        sum += (mask & (2 ** index) ? 1 : -1) * nonZero[index]!\n      }\n      const permuted = sum / totalPairs\n      if (isExtreme(permuted, statistic, config.alternative)) extreme++\n    }\n    return {\n      statistic,\n      pValue: extreme / assignments,\n      alternative: config.alternative,\n      method: 'exact',\n      assignments,\n      nClusters: clusters.length,\n      nNonZeroClusters: nonZero.length,\n      seed: null,\n    }\n  }\n\n  const signFlipSeed = (config.seed ^ SIGN_FLIP_SEED_SALT) >>> 0\n  const rng = mulberry32(signFlipSeed)\n  let extreme = 0\n  for (let draw = 0; draw < config.signFlipResamples; draw++) {\n    let sum = 0\n    for (const delta of nonZero) sum += (rng() < 0.5 ? -1 : 1) * delta\n    const permuted = sum / totalPairs\n    if (isExtreme(permuted, statistic, config.alternative)) extreme++\n  }\n  return {\n    statistic,\n    pValue: (extreme + 1) / (config.signFlipResamples + 1),\n    alternative: config.alternative,\n    method: 'monte-carlo',\n    assignments: config.signFlipResamples,\n    nClusters: clusters.length,\n    nNonZeroClusters: nonZero.length,\n    seed: signFlipSeed,\n  }\n}\n\nfunction isExtreme(\n  candidate: number,\n  observed: number,\n  alternative: ClusterSignFlipAlternative,\n): boolean {\n  const tolerance = Number.EPSILON * Math.max(1, Math.abs(observed)) * 16\n  if (alternative === 'greater') return candidate >= observed - tolerance\n  if (alternative === 'less') return candidate <= observed + tolerance\n  return Math.abs(candidate) >= Math.abs(observed) - tolerance\n}\n\nfunction mean(values: readonly number[]): number {\n  return values.reduce((sum, value) => sum + value, 0) / values.length\n}\n","/**\n * Experiment tracker — git-provenanced experiment log with N-rep stats and a\n * KEEP / REGRESSION / NOISE verdict against a parent.\n *\n * Every loop the fleet runs reduces to the same question: \"I ran the candidate\n * N times — is the median measurably better than the parent, or is the delta\n * inside the noise band?\" The hand-rolled copies bake a fixed score scale\n * (percentage points), a fixed store path (`.evolve/experiments-v2.json`), and\n * `execSync('git …')` straight into the module. This is the canonical version:\n * provenance and persistence are injected, thresholds are configurable, and the\n * stats + verdict are pure functions you can unit-test without a git repo or a\n * filesystem.\n *\n * Stats per experiment: median / mean / min / max / iqr / stddev / passRate /\n * n, plus a `stable` flag (`iqr < iqrUnstableAbove && stddev < stddevUnstableAbove`).\n *\n * Verdict against a parent (both must have `n >= minRepsForVerdict`):\n *   - NOISE       — the candidate is too unstable to judge (`!stable`)\n *   - KEEP        — `medianDelta >  keepThreshold`\n *   - REGRESSION  — `medianDelta < -regressionThreshold`\n *   - NOISE       — otherwise (delta inside the band)\n * With no parent (or insufficient reps) the verdict is the neutral ITERATE.\n */\n\nimport { execSync } from 'node:child_process'\nimport { iqr } from './baseline'\nimport { ValidationError } from './errors'\n\n/** Verdict for one experiment relative to its parent. ITERATE is the neutral\n *  \"keep collecting reps / no parent to compare against\" state. */\nexport type ExperimentVerdict = 'KEEP' | 'ITERATE' | 'NOISE' | 'REGRESSION'\n\n/** Git provenance for the working tree an experiment was run from. */\nexport interface ExperimentProvenance {\n  /** Commit sha (short or full — the tracker does not interpret it). */\n  commit: string\n  /** First line of the commit message. */\n  message: string\n  /** Files changed vs the parent commit, or a marker like 'uncommitted'. */\n  changedFiles: string[]\n}\n\n/** A single repetition of an experiment, carrying the score the verdict is\n *  computed on plus any free-form per-rep metrics the consumer wants kept. */\nexport interface ExperimentRep {\n  /** 0-indexed repetition number within the experiment. */\n  rep: number\n  /** The score this rep is judged on (same scale as the thresholds). */\n  score: number\n  /** ISO timestamp the rep completed. */\n  timestamp: string\n  /** Whether this rep passed the consumer's own gate — folded into `passRate`. */\n  passed?: boolean\n  /** Free-form numeric metrics retained for later analysis. */\n  metrics?: Record<string, number>\n}\n\nexport interface ExperimentStats {\n  median: number\n  mean: number\n  min: number\n  max: number\n  /** Inter-quartile range of the rep scores. */\n  iqr: number\n  /** Population standard deviation of the rep scores. */\n  stddev: number\n  /** Fraction of reps with `passed === true`, over reps that set `passed`.\n   *  null when no rep declared a pass/fail outcome. */\n  passRate: number | null\n  /** Number of reps. */\n  n: number\n  /** True when the sample is tight enough to trust for a verdict. */\n  stable: boolean\n}\n\nexport interface Experiment {\n  /** Stable id for the experiment. */\n  id: string\n  /** Free-form label / config descriptor. */\n  label: string\n  /** Git provenance captured when the experiment was created. */\n  provenance: ExperimentProvenance\n  /** Parent experiment id this candidate is compared against, if any. */\n  parentId?: string\n  /** One-line summary of what changed from the parent. */\n  changeSummary: string\n  reps: ExperimentRep[]\n  stats: ExperimentStats\n  verdict: ExperimentVerdict\n  /** ISO timestamp the experiment was created. */\n  createdAt: string\n}\n\nexport interface ImprovementThresholds {\n  /** medianDelta strictly above this ⇒ KEEP. Default 5. */\n  keepThreshold?: number\n  /** medianDelta strictly below the negative of this ⇒ REGRESSION. Default 5. */\n  regressionThreshold?: number\n  /** iqr at or above this ⇒ unstable. Default 10. */\n  iqrUnstableAbove?: number\n  /** stddev at or above this ⇒ unstable. Default Infinity (iqr-only stability). */\n  stddevUnstableAbove?: number\n  /** Reps required on BOTH candidate and parent before a verdict is rendered.\n   *  Default 3. */\n  minRepsForVerdict?: number\n}\n\nexport interface ImprovementVerdictResult {\n  verdict: ExperimentVerdict\n  /** candidate.median − parent.median; null when no parent or insufficient reps. */\n  medianDelta: number | null\n  /** Human-readable reason for the verdict — for dashboards and logs. */\n  reason: string\n}\n\nconst DEFAULTS: Required<ImprovementThresholds> = {\n  keepThreshold: 5,\n  regressionThreshold: 5,\n  iqrUnstableAbove: 10,\n  stddevUnstableAbove: Number.POSITIVE_INFINITY,\n  minRepsForVerdict: 3,\n}\n\nfunction resolveThresholds(t: ImprovementThresholds | undefined): Required<ImprovementThresholds> {\n  const r = { ...DEFAULTS, ...(t ?? {}) }\n  if (r.keepThreshold < 0) {\n    throw new ValidationError(\n      `experiment-tracker: keepThreshold must be >= 0, got ${r.keepThreshold}`,\n    )\n  }\n  if (r.regressionThreshold < 0) {\n    throw new ValidationError(\n      `experiment-tracker: regressionThreshold must be >= 0, got ${r.regressionThreshold}`,\n    )\n  }\n  if (r.minRepsForVerdict < 1) {\n    throw new ValidationError(\n      `experiment-tracker: minRepsForVerdict must be >= 1, got ${r.minRepsForVerdict}`,\n    )\n  }\n  return r\n}\n\nfunction median(sorted: number[]): number {\n  const n = sorted.length\n  if (n === 0) return 0\n  const mid = Math.floor(n / 2)\n  return n % 2 === 0 ? (sorted[mid - 1]! + sorted[mid]!) / 2 : sorted[mid]!\n}\n\n/** Population standard deviation (÷n). 0 for fewer than 2 values. */\nfunction stddev(values: number[], mean: number): number {\n  if (values.length < 2) return 0\n  const variance = values.reduce((acc, v) => acc + (v - mean) ** 2, 0) / values.length\n  return Math.sqrt(variance)\n}\n\n/**\n * Compute the N-rep statistics for a set of reps. Pure — no I/O. The `stable`\n * flag is the trust gate the verdict depends on: a sample whose spread exceeds\n * the configured bounds can't distinguish a real delta from run-to-run noise.\n */\nexport function computeExperimentStats(\n  reps: ExperimentRep[],\n  thresholds?: ImprovementThresholds,\n): ExperimentStats {\n  const t = resolveThresholds(thresholds)\n  const n = reps.length\n  if (n === 0) {\n    return {\n      median: 0,\n      mean: 0,\n      min: 0,\n      max: 0,\n      iqr: 0,\n      stddev: 0,\n      passRate: null,\n      n: 0,\n      stable: false,\n    }\n  }\n  const scores = reps.map((r) => {\n    if (!Number.isFinite(r.score)) {\n      throw new ValidationError(`experiment-tracker: rep ${r.rep} has non-finite score ${r.score}`)\n    }\n    return r.score\n  })\n  const sorted = [...scores].sort((a, b) => a - b)\n  const mean = scores.reduce((s, v) => s + v, 0) / n\n  const sd = stddev(scores, mean)\n  const spread = iqr(scores)\n  const rated = reps.filter((r) => typeof r.passed === 'boolean')\n  const passRate = rated.length === 0 ? null : rated.filter((r) => r.passed).length / rated.length\n  const stable = spread < t.iqrUnstableAbove && sd < t.stddevUnstableAbove\n  return {\n    median: median(sorted),\n    mean,\n    min: sorted[0]!,\n    max: sorted[n - 1]!,\n    iqr: spread,\n    stddev: sd,\n    passRate,\n    n,\n    stable,\n  }\n}\n\n/**\n * Verdict for a candidate against its parent. Pure — operates on already-computed\n * stats. KEEP/REGRESSION require both sides to have `>= minRepsForVerdict` reps\n * AND the candidate to be `stable`; otherwise the result is NOISE (unstable) or\n * ITERATE (not enough reps / no parent).\n */\nexport function improvementVerdict(\n  candidate: ExperimentStats,\n  parent: ExperimentStats | null,\n  thresholds?: ImprovementThresholds,\n): ImprovementVerdictResult {\n  const t = resolveThresholds(thresholds)\n  if (!parent) {\n    return {\n      verdict: 'ITERATE',\n      medianDelta: null,\n      reason: 'no parent experiment to compare against',\n    }\n  }\n  if (candidate.n < t.minRepsForVerdict || parent.n < t.minRepsForVerdict) {\n    return {\n      verdict: 'ITERATE',\n      medianDelta: null,\n      reason: `need >= ${t.minRepsForVerdict} reps on both sides (candidate n=${candidate.n}, parent n=${parent.n})`,\n    }\n  }\n  if (!candidate.stable) {\n    return {\n      verdict: 'NOISE',\n      medianDelta: candidate.median - parent.median,\n      reason: `candidate unstable (iqr=${candidate.iqr}, stddev=${candidate.stddev.toFixed(2)})`,\n    }\n  }\n  const medianDelta = candidate.median - parent.median\n  if (medianDelta > t.keepThreshold) {\n    return { verdict: 'KEEP', medianDelta, reason: `median +${medianDelta} > +${t.keepThreshold}` }\n  }\n  if (medianDelta < -t.regressionThreshold) {\n    return {\n      verdict: 'REGRESSION',\n      medianDelta,\n      reason: `median ${medianDelta} < -${t.regressionThreshold}`,\n    }\n  }\n  return {\n    verdict: 'NOISE',\n    medianDelta,\n    reason: `median delta ${medianDelta} inside noise band [-${t.regressionThreshold}, +${t.keepThreshold}]`,\n  }\n}\n\n// ── Provenance + persistence seams ───────────────────────────────────\n\n/** Reads git provenance for the working tree. Inject a fake in tests; the\n *  default implementation shells out to `git`. */\nexport type ProvenanceReader = () => ExperimentProvenance | Promise<ExperimentProvenance>\n\n/** Persistence seam for the experiment log. Inject in-memory in tests; the\n *  filesystem implementation is `fileExperimentStore`. */\nexport interface ExperimentStore {\n  load(): Promise<Experiment[]>\n  save(experiments: Experiment[]): Promise<void>\n}\n\n/**\n * Default provenance reader: `git rev-parse HEAD`, the subject line, and the\n * files changed vs `HEAD~1`. Fail-loud — a tracker that silently logs\n * `commit: 'unknown'` corrupts the provenance the whole point of the log is to\n * carry. When the working tree genuinely has no parent commit, pass an override.\n */\nexport const gitProvenanceReader: ProvenanceReader = () => {\n  const run = (cmd: string): string => execSync(cmd, { encoding: 'utf8' }).trim()\n  const commit = run('git rev-parse --short HEAD')\n  const message = run('git log -1 --format=%s')\n  const changedRaw = run('git diff --name-only HEAD~1')\n  const changedFiles = changedRaw.length === 0 ? [] : changedRaw.split('\\n').filter(Boolean)\n  return { commit, message, changedFiles }\n}\n\n/** In-memory store — the default when no persistence is wanted (tests, ephemeral\n *  runs). State lives on the instance. */\nexport function inMemoryExperimentStore(initial: Experiment[] = []): ExperimentStore {\n  let state = initial.map((e) => structuredClone(e))\n  return {\n    async load() {\n      return state.map((e) => structuredClone(e))\n    },\n    async save(experiments) {\n      state = experiments.map((e) => structuredClone(e))\n    },\n  }\n}\n\n/** Filesystem store — a single JSON array at `path`, created on first save. */\nexport function fileExperimentStore(path: string): ExperimentStore {\n  return {\n    async load() {\n      const fs = await import('node:fs/promises')\n      try {\n        const raw = await fs.readFile(path, 'utf8')\n        const parsed = JSON.parse(raw)\n        if (!Array.isArray(parsed)) {\n          throw new ValidationError(`experiment-tracker: store at ${path} is not a JSON array`)\n        }\n        return parsed as Experiment[]\n      } catch (err) {\n        if ((err as NodeJS.ErrnoException).code === 'ENOENT') return []\n        throw err\n      }\n    },\n    async save(experiments) {\n      const fs = await import('node:fs/promises')\n      const pathMod = await import('node:path')\n      await fs.mkdir(pathMod.dirname(path), { recursive: true })\n      await fs.writeFile(path, JSON.stringify(experiments, null, 2), 'utf8')\n    },\n  }\n}\n\nexport interface ExperimentTrackerOptions {\n  store?: ExperimentStore\n  provenanceReader?: ProvenanceReader\n  thresholds?: ImprovementThresholds\n  /** Clock seam for deterministic timestamps in tests. Default `Date.now`. */\n  now?: () => number\n}\n\nexport interface CreateExperimentInput {\n  id: string\n  label: string\n  changeSummary: string\n  parentId?: string\n  /** Override provenance instead of reading from git (e.g. CI metadata). */\n  provenance?: ExperimentProvenance\n}\n\n/**\n * Stateful tracker over an `ExperimentStore`. Create an experiment (provenance\n * is captured once), append reps as they complete (stats + verdict recompute on\n * every append), and read the log back for a dashboard. All persistence and git\n * access flow through the injected seams, so the tracker is fully testable\n * without a repo or disk.\n */\nexport class ExperimentTracker {\n  private readonly store: ExperimentStore\n  private readonly provenanceReader: ProvenanceReader\n  private readonly thresholds: Required<ImprovementThresholds>\n  private readonly now: () => number\n\n  constructor(options: ExperimentTrackerOptions = {}) {\n    this.store = options.store ?? inMemoryExperimentStore()\n    this.provenanceReader = options.provenanceReader ?? gitProvenanceReader\n    this.thresholds = resolveThresholds(options.thresholds)\n    this.now = options.now ?? Date.now\n  }\n\n  async create(input: CreateExperimentInput): Promise<Experiment> {\n    const experiments = await this.store.load()\n    if (experiments.some((e) => e.id === input.id)) {\n      throw new ValidationError(`experiment-tracker: experiment id \"${input.id}\" already exists`)\n    }\n    if (input.parentId && !experiments.some((e) => e.id === input.parentId)) {\n      throw new ValidationError(\n        `experiment-tracker: parent experiment \"${input.parentId}\" not found`,\n      )\n    }\n    const provenance = input.provenance ?? (await this.provenanceReader())\n    const experiment: Experiment = {\n      id: input.id,\n      label: input.label,\n      provenance,\n      parentId: input.parentId,\n      changeSummary: input.changeSummary,\n      reps: [],\n      stats: computeExperimentStats([], this.thresholds),\n      verdict: 'ITERATE',\n      createdAt: new Date(this.now()).toISOString(),\n    }\n    experiments.push(experiment)\n    await this.store.save(experiments)\n    return structuredClone(experiment)\n  }\n\n  /** Append a rep (its `rep` index defaults to the current rep count) and\n   *  recompute stats + verdict. Returns the updated experiment. */\n  async addRep(\n    experimentId: string,\n    rep: Omit<ExperimentRep, 'rep' | 'timestamp'> & { rep?: number; timestamp?: string },\n  ): Promise<Experiment> {\n    const experiments = await this.store.load()\n    const exp = experiments.find((e) => e.id === experimentId)\n    if (!exp)\n      throw new ValidationError(`experiment-tracker: experiment \"${experimentId}\" not found`)\n    const fullRep: ExperimentRep = {\n      rep: rep.rep ?? exp.reps.length,\n      score: rep.score,\n      passed: rep.passed,\n      metrics: rep.metrics,\n      timestamp: rep.timestamp ?? new Date(this.now()).toISOString(),\n    }\n    exp.reps.push(fullRep)\n    exp.stats = computeExperimentStats(exp.reps, this.thresholds)\n    const parent = exp.parentId ? experiments.find((e) => e.id === exp.parentId) : undefined\n    exp.verdict = improvementVerdict(exp.stats, parent?.stats ?? null, this.thresholds).verdict\n    await this.store.save(experiments)\n    return structuredClone(exp)\n  }\n\n  async get(experimentId: string): Promise<Experiment | undefined> {\n    const experiments = await this.store.load()\n    const found = experiments.find((e) => e.id === experimentId)\n    return found ? structuredClone(found) : undefined\n  }\n\n  async list(): Promise<Experiment[]> {\n    return this.store.load()\n  }\n\n  /** Full verdict (not just the enum) for an experiment vs its parent. */\n  async verdictFor(experimentId: string): Promise<ImprovementVerdictResult> {\n    const experiments = await this.store.load()\n    const exp = experiments.find((e) => e.id === experimentId)\n    if (!exp)\n      throw new ValidationError(`experiment-tracker: experiment \"${experimentId}\" not found`)\n    const parent = exp.parentId ? experiments.find((e) => e.id === exp.parentId) : undefined\n    return improvementVerdict(exp.stats, parent?.stats ?? null, this.thresholds)\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;AA+HA,MAAM,8BAA8B;AACpC,MAAM,8BAA8B;AACpC,MAAM,gBAAgB;AACtB,MAAM,8BAA8B;AACpC,MAAM,sBAAsB;;;;;;;AAQ5B,SAAgB,sBACd,MACA,SACmC;CACnC,MAAM,SAAS,gBAAgB,OAAO;CAEtC,MAAM,SAAS,SADG,oBAAoB,MAAM,OACZ,GAAG;EACjC,aAAa,QAAQ;EACrB,cAAc,QAAQ;CACxB,CAAC;CAED,MAAM,eAAe,OAAO,MAAM,KAAK,SAAqC;EAC1E,MAAM,WAAW,KAAK;EACtB,MAAM,YAAY,KAAK;EACvB,IAAI,SAAS,eAAe,UAAU,YACpC,MAAM,IAAI,gBACR,mCAAmC,KAAK,QAAQ,QAAQ,KAAK,SAAS,sBAC/D,SAAS,WAAW,QAAQ,UAAU,WAAW,GAC1D;EAEF,OAAO;GACL,SAAS,KAAK;GACd,UAAU,KAAK;GACf,YAAY,SAAS;GACrB,UAAU,SAAS;GACnB,WAAW,UAAU;GACrB,cAAc,SAAS;GACvB,eAAe,UAAU;EAC3B;CACF,CAAC;CAED,MAAM,mBAAmB,OAAO,iBAAiB,KAC9C,QAAS,IAA2B,QACvC;CACA,MAAM,oBAAoB,OAAO,kBAAkB,KAChD,QAAS,IAA2B,QACvC;CAEA,IAAI,aAAa,WAAW,GAC1B,OAAO;EAAE;EAAc;EAAkB;EAAmB,YAAY;CAAK;CAG/E,MAAM,WAAW,kBAAkB,YAAY;CAC/C,MAAM,MAAM,SAAS,QAAQ,KAAK,YAAY,MAAM,QAAQ,KAAK,CAAC;CAClE,MAAM,MAAM,SAAS,QAAQ,KAAK,YAAY,MAAM,QAAQ,KAAK,CAAC;CAClE,MAAM,8BAA8B,MAAM,OAAO,aAAa;CAC9D,MAAM,mBAAmB,KAAK,SAAS,KAAK,YAAY,QAAQ,cAAc,CAAC;CAC/E,MAAM,YAAY,SAAS,SAAS,IAAI,OAAO,iBAAiB,UAAU,MAAM;CAChF,MAAM,WAAW,gBAAgB,UAAU,MAAM;CAEjD,OAAO;EACL;EACA;EACA;EACA,YAAY;GACV,QAAQ,aAAa;GACrB,WAAW,SAAS;GACpB;GACA;GACA;GACA;GACA;GACA;GACA;EACF;CACF;AACF;AAWA,SAAS,gBAAsB,SAA8D;CAC3F,qBAAqB,eAAe,QAAQ,WAAW;CACvD,qBAAqB,gBAAgB,QAAQ,YAAY;CACzD,IAAI,QAAQ,gBAAgB,QAAQ,cAClC,MAAM,IAAI,gBACR,iEAAiE,QAAQ,YAAY,EACvF;CAEF,IAAI,CAAC,OAAO,UAAU,QAAQ,IAAI,GAChC,MAAM,IAAI,gBAAgB,uDAAuD,QAAQ,MAAM;CAEjG,MAAM,aAAa,QAAQ,cAAc;CACzC,IAAI,CAAC,OAAO,SAAS,UAAU,KAAK,cAAc,KAAK,cAAc,GACnE,MAAM,IAAI,gBACR,2DAA2D,YAC7D;CAEF,MAAM,qBAAqB,QAAQ,sBAAsB;CACzD,oBAAoB,sBAAsB,kBAAkB;CAC5D,MAAM,+BAA+B,KAAK,IAAI;CAC9C,MAAM,4BAA4B,KAAK,KACrC,+BAA+B,OAAO,UAAU,KAAK,IAAI,GAAG,4BAA4B,IAAI,CAC9F;CACA,IAAI,qBAAqB,2BACvB,MAAM,IAAI,gBACR,8DAA8D,0BAA0B,kBAAkB,WAAW,+CAA+C,oBACtK;CAEF,MAAM,oBAAoB,QAAQ,qBAAqB;CACvD,oBAAoB,qBAAqB,iBAAiB;CAC1D,MAAM,oBAAoB,QAAQ,qBAAqB;CACvD,IACE,CAAC,OAAO,UAAU,iBAAiB,KACnC,oBAAoB,KACpB,oBAAoB,6BAEpB,MAAM,IAAI,gBACR,qEAAqE,4BAA4B,SAAS,mBAC5G;CAEF,MAAM,cAAc,QAAQ,eAAe;CAC3C,IAAI,gBAAgB,eAAe,gBAAgB,aAAa,gBAAgB,QAC9E,MAAM,IAAI,gBACR,qFAAqF,OAAO,WAAW,GACzG;CAEF,OAAO;EACL,MAAM,QAAQ;EACd;EACA;EACA;EACA;EACA;CACF;AACF;AAEA,SAAS,oBAAoB,MAAc,OAAqB;CAC9D,IAAI,CAAC,OAAO,UAAU,KAAK,KAAK,SAAS,GACvC,MAAM,IAAI,gBACR,0BAA0B,KAAK,mCAAmC,OACpE;CAEF,IAAI,QAAQ,eACV,MAAM,IAAI,gBACR,0BAA0B,KAAK,mBAAmB,cAAc,QAAQ,OAC1E;AAEJ;AAEA,SAAS,oBACP,MACA,SACsB;CACtB,MAAM,YAAkC,CAAC;CACzC,KAAK,MAAM,YAAY,MAAM;EAC3B,MAAM,MAAM,QAAQ,IAAI,QAAQ;EAChC,qBAAqB,OAAO,GAAG;EAC/B,IAAI,QAAQ,QAAQ,eAAe,QAAQ,QAAQ,cAAc;EAEjE,MAAM,UAAU,QAAQ,QAAQ,QAAQ;EACxC,MAAM,aAAa,QAAQ,WAAW,QAAQ;EAC9C,MAAM,OAAO,QAAQ,KAAK,QAAQ;EAClC,MAAM,SAAS,QAAQ,SAAS,QAAQ;EACxC,qBAAqB,WAAW,OAAO;EACvC,qBAAqB,cAAc,UAAU;EAC7C,IAAI,OAAO,SAAS,WAClB,MAAM,IAAI,gBACR,yEAAyE,QAAQ,EACnF;EAEF,IAAI,WAAW,KAAA,GAAW,qBAAqB,UAAU,MAAM;EAC/D,UAAU,KAAK;GAAE;GAAS;GAAY;GAAK;GAAM;GAAQ;EAAS,CAAC;CACrE;CACA,OAAO;AACT;AAEA,SAAS,qBAAqB,MAAc,OAAqB;CAC/D,IAAI,OAAO,UAAU,YAAY,MAAM,KAAK,CAAC,CAAC,WAAW,GACvD,MAAM,IAAI,gBACR,0BAA0B,KAAK,yCACjC;AAEJ;AAEA,SAAS,kBACP,OAC0B;CAC1B,MAAM,4BAAY,IAAI,IAA0D;CAChF,KAAK,MAAM,QAAQ,OAAO;EACxB,MAAM,UAAU,UAAU,IAAI,KAAK,UAAU,KAAK;GAAE,QAAQ;GAAG,KAAK;GAAG,KAAK;EAAE;EAC9E,QAAQ;EACR,IAAI,KAAK,iBAAiB,CAAC,KAAK,cAAc,QAAQ;OACjD,IAAI,KAAK,gBAAgB,CAAC,KAAK,eAAe,QAAQ;EAC3D,UAAU,IAAI,KAAK,YAAY,OAAO;CACxC;CACA,OAAO,CAAC,GAAG,UAAU,QAAQ,CAAC,CAAC,CAC5B,MAAM,CAAC,IAAI,CAAC,OAAQ,IAAI,IAAI,KAAK,IAAI,IAAI,IAAI,CAAE,CAAC,CAChD,KAAK,CAAC,YAAY,cAAc;EAC/B;EACA,GAAG;EACH,iBAAiB,QAAQ,MAAM,QAAQ,OAAO,QAAQ;CACxD,EAAE;AACN;AAEA,SAAS,iBACP,UACA,QAC0B;CAC1B,MAAM,MAAM,WAAW,OAAO,IAAI;CAClC,MAAM,UAAU,IAAI,MAAc,OAAO,kBAAkB;CAC3D,KAAK,IAAI,OAAO,GAAG,OAAO,OAAO,oBAAoB,QAAQ;EAC3D,IAAI,gBAAgB;EACpB,IAAI,YAAY;EAChB,KAAK,IAAI,QAAQ,GAAG,QAAQ,SAAS,QAAQ,SAAS;GACpD,MAAM,UAAU,SAAS,KAAK,MAAM,IAAI,IAAI,SAAS,MAAM;GAC3D,iBAAiB,QAAQ,MAAM,QAAQ;GACvC,aAAa,QAAQ;EACvB;EACA,QAAQ,QAAQ,gBAAgB;CAClC;CACA,QAAQ,MAAM,GAAG,MAAM,IAAI,CAAC;CAC5B,MAAM,QAAQ,IAAI,OAAO;CACzB,MAAM,aAAa,KAAK,MAAO,QAAQ,IAAK,OAAO,kBAAkB;CACrE,MAAM,aAAa,KAAK,IACtB,OAAO,qBAAqB,GAC5B,KAAK,MAAM,IAAI,QAAQ,KAAK,OAAO,kBAAkB,IAAI,CAC3D;CACA,OAAO;EACL,WAAW;EACX,OAAO,QAAQ;EACf,OAAO,QAAQ,KAAK,IAAI,YAAY,UAAU;EAC9C,YAAY,OAAO;EACnB,WAAW,OAAO;EAClB,MAAM,OAAO;CACf;AACF;AAEA,SAAS,gBACP,UACA,QACuB;CACvB,MAAM,gBAAgB,SAAS,KAAK,YAAY,QAAQ,MAAM,QAAQ,GAAG;CACzE,MAAM,UAAU,cAAc,QAAQ,UAAU,UAAU,CAAC;CAC3D,MAAM,aAAa,SAAS,QAAQ,KAAK,YAAY,MAAM,QAAQ,QAAQ,CAAC;CAC5E,MAAM,YAAY,cAAc,QAAQ,KAAK,UAAU,MAAM,OAAO,CAAC,IAAI;CACzE,IAAI,QAAQ,UAAU,OAAO,mBAAmB;EAC9C,MAAM,cAAc,KAAK,QAAQ;EACjC,IAAI,UAAU;EACd,KAAK,IAAI,OAAO,GAAG,OAAO,aAAa,QAAQ;GAC7C,IAAI,MAAM;GACV,KAAK,IAAI,QAAQ,GAAG,QAAQ,QAAQ,QAAQ,SAC1C,QAAQ,OAAQ,KAAK,QAAS,IAAI,MAAM,QAAQ;GAGlD,IAAI,UADa,MAAM,YACC,WAAW,OAAO,WAAW,GAAG;EAC1D;EACA,OAAO;GACL;GACA,QAAQ,UAAU;GAClB,aAAa,OAAO;GACpB,QAAQ;GACR;GACA,WAAW,SAAS;GACpB,kBAAkB,QAAQ;GAC1B,MAAM;EACR;CACF;CAEA,MAAM,gBAAgB,OAAO,OAAO,yBAAyB;CAC7D,MAAM,MAAM,WAAW,YAAY;CACnC,IAAI,UAAU;CACd,KAAK,IAAI,OAAO,GAAG,OAAO,OAAO,mBAAmB,QAAQ;EAC1D,IAAI,MAAM;EACV,KAAK,MAAM,SAAS,SAAS,QAAQ,IAAI,IAAI,KAAM,KAAK,KAAK;EAE7D,IAAI,UADa,MAAM,YACC,WAAW,OAAO,WAAW,GAAG;CAC1D;CACA,OAAO;EACL;EACA,SAAS,UAAU,MAAM,OAAO,oBAAoB;EACpD,aAAa,OAAO;EACpB,QAAQ;EACR,aAAa,OAAO;EACpB,WAAW,SAAS;EACpB,kBAAkB,QAAQ;EAC1B,MAAM;CACR;AACF;AAEA,SAAS,UACP,WACA,UACA,aACS;CACT,MAAM,YAAY,OAAO,UAAU,KAAK,IAAI,GAAG,KAAK,IAAI,QAAQ,CAAC,IAAI;CACrE,IAAI,gBAAgB,WAAW,OAAO,aAAa,WAAW;CAC9D,IAAI,gBAAgB,QAAQ,OAAO,aAAa,WAAW;CAC3D,OAAO,KAAK,IAAI,SAAS,KAAK,KAAK,IAAI,QAAQ,IAAI;AACrD;AAEA,SAAS,KAAK,QAAmC;CAC/C,OAAO,OAAO,QAAQ,KAAK,UAAU,MAAM,OAAO,CAAC,IAAI,OAAO;AAChE;;;;;;;;;;;;;;;;;;;;;;;;;;ACnUA,MAAM,WAA4C;CAChD,eAAe;CACf,qBAAqB;CACrB,kBAAkB;CAClB,qBAAqB,OAAO;CAC5B,mBAAmB;AACrB;AAEA,SAAS,kBAAkB,GAAuE;CAChG,MAAM,IAAI;EAAE,GAAG;EAAU,GAAI,KAAK,CAAC;CAAG;CACtC,IAAI,EAAE,gBAAgB,GACpB,MAAM,IAAI,gBACR,uDAAuD,EAAE,eAC3D;CAEF,IAAI,EAAE,sBAAsB,GAC1B,MAAM,IAAI,gBACR,6DAA6D,EAAE,qBACjE;CAEF,IAAI,EAAE,oBAAoB,GACxB,MAAM,IAAI,gBACR,2DAA2D,EAAE,mBAC/D;CAEF,OAAO;AACT;AAEA,SAAS,OAAO,QAA0B;CACxC,MAAM,IAAI,OAAO;CACjB,IAAI,MAAM,GAAG,OAAO;CACpB,MAAM,MAAM,KAAK,MAAM,IAAI,CAAC;CAC5B,OAAO,IAAI,MAAM,KAAK,OAAO,MAAM,KAAM,OAAO,QAAS,IAAI,OAAO;AACtE;;AAGA,SAAS,OAAO,QAAkB,MAAsB;CACtD,IAAI,OAAO,SAAS,GAAG,OAAO;CAC9B,MAAM,WAAW,OAAO,QAAQ,KAAK,MAAM,OAAO,IAAI,SAAS,GAAG,CAAC,IAAI,OAAO;CAC9E,OAAO,KAAK,KAAK,QAAQ;AAC3B;;;;;;AAOA,SAAgB,uBACd,MACA,YACiB;CACjB,MAAM,IAAI,kBAAkB,UAAU;CACtC,MAAM,IAAI,KAAK;CACf,IAAI,MAAM,GACR,OAAO;EACL,QAAQ;EACR,MAAM;EACN,KAAK;EACL,KAAK;EACL,KAAK;EACL,QAAQ;EACR,UAAU;EACV,GAAG;EACH,QAAQ;CACV;CAEF,MAAM,SAAS,KAAK,KAAK,MAAM;EAC7B,IAAI,CAAC,OAAO,SAAS,EAAE,KAAK,GAC1B,MAAM,IAAI,gBAAgB,2BAA2B,EAAE,IAAI,wBAAwB,EAAE,OAAO;EAE9F,OAAO,EAAE;CACX,CAAC;CACD,MAAM,SAAS,CAAC,GAAG,MAAM,CAAC,CAAC,MAAM,GAAG,MAAM,IAAI,CAAC;CAC/C,MAAM,OAAO,OAAO,QAAQ,GAAG,MAAM,IAAI,GAAG,CAAC,IAAI;CACjD,MAAM,KAAK,OAAO,QAAQ,IAAI;CAC9B,MAAM,SAAS,IAAI,MAAM;CACzB,MAAM,QAAQ,KAAK,QAAQ,MAAM,OAAO,EAAE,WAAW,SAAS;CAC9D,MAAM,WAAW,MAAM,WAAW,IAAI,OAAO,MAAM,QAAQ,MAAM,EAAE,MAAM,CAAC,CAAC,SAAS,MAAM;CAC1F,MAAM,SAAS,SAAS,EAAE,oBAAoB,KAAK,EAAE;CACrD,OAAO;EACL,QAAQ,OAAO,MAAM;EACrB;EACA,KAAK,OAAO;EACZ,KAAK,OAAO,IAAI;EAChB,KAAK;EACL,QAAQ;EACR;EACA;EACA;CACF;AACF;;;;;;;AAQA,SAAgB,mBACd,WACA,QACA,YAC0B;CAC1B,MAAM,IAAI,kBAAkB,UAAU;CACtC,IAAI,CAAC,QACH,OAAO;EACL,SAAS;EACT,aAAa;EACb,QAAQ;CACV;CAEF,IAAI,UAAU,IAAI,EAAE,qBAAqB,OAAO,IAAI,EAAE,mBACpD,OAAO;EACL,SAAS;EACT,aAAa;EACb,QAAQ,WAAW,EAAE,kBAAkB,mCAAmC,UAAU,EAAE,aAAa,OAAO,EAAE;CAC9G;CAEF,IAAI,CAAC,UAAU,QACb,OAAO;EACL,SAAS;EACT,aAAa,UAAU,SAAS,OAAO;EACvC,QAAQ,2BAA2B,UAAU,IAAI,WAAW,UAAU,OAAO,QAAQ,CAAC,EAAE;CAC1F;CAEF,MAAM,cAAc,UAAU,SAAS,OAAO;CAC9C,IAAI,cAAc,EAAE,eAClB,OAAO;EAAE,SAAS;EAAQ;EAAa,QAAQ,WAAW,YAAY,MAAM,EAAE;CAAgB;CAEhG,IAAI,cAAc,CAAC,EAAE,qBACnB,OAAO;EACL,SAAS;EACT;EACA,QAAQ,UAAU,YAAY,MAAM,EAAE;CACxC;CAEF,OAAO;EACL,SAAS;EACT;EACA,QAAQ,gBAAgB,YAAY,uBAAuB,EAAE,oBAAoB,KAAK,EAAE,cAAc;CACxG;AACF;;;;;;;AAqBA,MAAa,4BAA8C;CACzD,MAAM,OAAO,QAAwB,SAAS,KAAK,EAAE,UAAU,OAAO,CAAC,CAAC,CAAC,KAAK;CAC9E,MAAM,SAAS,IAAI,4BAA4B;CAC/C,MAAM,UAAU,IAAI,wBAAwB;CAC5C,MAAM,aAAa,IAAI,6BAA6B;CAEpD,OAAO;EAAE;EAAQ;EAAS,cADL,WAAW,WAAW,IAAI,CAAC,IAAI,WAAW,MAAM,IAAI,CAAC,CAAC,OAAO,OAAO;CAClD;AACzC;;;AAIA,SAAgB,wBAAwB,UAAwB,CAAC,GAAoB;CACnF,IAAI,QAAQ,QAAQ,KAAK,MAAM,gBAAgB,CAAC,CAAC;CACjD,OAAO;EACL,MAAM,OAAO;GACX,OAAO,MAAM,KAAK,MAAM,gBAAgB,CAAC,CAAC;EAC5C;EACA,MAAM,KAAK,aAAa;GACtB,QAAQ,YAAY,KAAK,MAAM,gBAAgB,CAAC,CAAC;EACnD;CACF;AACF;;AAGA,SAAgB,oBAAoB,MAA+B;CACjE,OAAO;EACL,MAAM,OAAO;GACX,MAAM,KAAK,MAAM,OAAO;GACxB,IAAI;IACF,MAAM,MAAM,MAAM,GAAG,SAAS,MAAM,MAAM;IAC1C,MAAM,SAAS,KAAK,MAAM,GAAG;IAC7B,IAAI,CAAC,MAAM,QAAQ,MAAM,GACvB,MAAM,IAAI,gBAAgB,gCAAgC,KAAK,qBAAqB;IAEtF,OAAO;GACT,SAAS,KAAK;IACZ,IAAK,IAA8B,SAAS,UAAU,OAAO,CAAC;IAC9D,MAAM;GACR;EACF;EACA,MAAM,KAAK,aAAa;GACtB,MAAM,KAAK,MAAM,OAAO;GACxB,MAAM,UAAU,MAAM,OAAO;GAC7B,MAAM,GAAG,MAAM,QAAQ,QAAQ,IAAI,GAAG,EAAE,WAAW,KAAK,CAAC;GACzD,MAAM,GAAG,UAAU,MAAM,KAAK,UAAU,aAAa,MAAM,CAAC,GAAG,MAAM;EACvE;CACF;AACF;;;;;;;;AA0BA,IAAa,oBAAb,MAA+B;CAC7B;CACA;CACA;CACA;CAEA,YAAY,UAAoC,CAAC,GAAG;EAClD,KAAK,QAAQ,QAAQ,SAAS,wBAAwB;EACtD,KAAK,mBAAmB,QAAQ,oBAAoB;EACpD,KAAK,aAAa,kBAAkB,QAAQ,UAAU;EACtD,KAAK,MAAM,QAAQ,OAAO,KAAK;CACjC;CAEA,MAAM,OAAO,OAAmD;EAC9D,MAAM,cAAc,MAAM,KAAK,MAAM,KAAK;EAC1C,IAAI,YAAY,MAAM,MAAM,EAAE,OAAO,MAAM,EAAE,GAC3C,MAAM,IAAI,gBAAgB,sCAAsC,MAAM,GAAG,iBAAiB;EAE5F,IAAI,MAAM,YAAY,CAAC,YAAY,MAAM,MAAM,EAAE,OAAO,MAAM,QAAQ,GACpE,MAAM,IAAI,gBACR,0CAA0C,MAAM,SAAS,YAC3D;EAEF,MAAM,aAAa,MAAM,cAAe,MAAM,KAAK,iBAAiB;EACpE,MAAM,aAAyB;GAC7B,IAAI,MAAM;GACV,OAAO,MAAM;GACb;GACA,UAAU,MAAM;GAChB,eAAe,MAAM;GACrB,MAAM,CAAC;GACP,OAAO,uBAAuB,CAAC,GAAG,KAAK,UAAU;GACjD,SAAS;GACT,WAAW,IAAI,KAAK,KAAK,IAAI,CAAC,CAAC,CAAC,YAAY;EAC9C;EACA,YAAY,KAAK,UAAU;EAC3B,MAAM,KAAK,MAAM,KAAK,WAAW;EACjC,OAAO,gBAAgB,UAAU;CACnC;;;CAIA,MAAM,OACJ,cACA,KACqB;EACrB,MAAM,cAAc,MAAM,KAAK,MAAM,KAAK;EAC1C,MAAM,MAAM,YAAY,MAAM,MAAM,EAAE,OAAO,YAAY;EACzD,IAAI,CAAC,KACH,MAAM,IAAI,gBAAgB,mCAAmC,aAAa,YAAY;EACxF,MAAM,UAAyB;GAC7B,KAAK,IAAI,OAAO,IAAI,KAAK;GACzB,OAAO,IAAI;GACX,QAAQ,IAAI;GACZ,SAAS,IAAI;GACb,WAAW,IAAI,aAAa,IAAI,KAAK,KAAK,IAAI,CAAC,CAAC,CAAC,YAAY;EAC/D;EACA,IAAI,KAAK,KAAK,OAAO;EACrB,IAAI,QAAQ,uBAAuB,IAAI,MAAM,KAAK,UAAU;EAC5D,MAAM,SAAS,IAAI,WAAW,YAAY,MAAM,MAAM,EAAE,OAAO,IAAI,QAAQ,IAAI,KAAA;EAC/E,IAAI,UAAU,mBAAmB,IAAI,OAAO,QAAQ,SAAS,MAAM,KAAK,UAAU,CAAC,CAAC;EACpF,MAAM,KAAK,MAAM,KAAK,WAAW;EACjC,OAAO,gBAAgB,GAAG;CAC5B;CAEA,MAAM,IAAI,cAAuD;EAE/D,MAAM,SAAQ,MADY,KAAK,MAAM,KAAK,EAAA,CAChB,MAAM,MAAM,EAAE,OAAO,YAAY;EAC3D,OAAO,QAAQ,gBAAgB,KAAK,IAAI,KAAA;CAC1C;CAEA,MAAM,OAA8B;EAClC,OAAO,KAAK,MAAM,KAAK;CACzB;;CAGA,MAAM,WAAW,cAAyD;EACxE,MAAM,cAAc,MAAM,KAAK,MAAM,KAAK;EAC1C,MAAM,MAAM,YAAY,MAAM,MAAM,EAAE,OAAO,YAAY;EACzD,IAAI,CAAC,KACH,MAAM,IAAI,gBAAgB,mCAAmC,aAAa,YAAY;EACxF,MAAM,SAAS,IAAI,WAAW,YAAY,MAAM,MAAM,EAAE,OAAO,IAAI,QAAQ,IAAI,KAAA;EAC/E,OAAO,mBAAmB,IAAI,OAAO,QAAQ,SAAS,MAAM,KAAK,UAAU;CAC7E;AACF"}

package/dist/experiment-tracker-IMntXr6J.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"experiment-tracker-IMntXr6J.d.ts","names":[],"sources":["../src/clustered-paired-binary.ts","../src/experiment-tracker.ts"],"mappings":";;;;;;;;;KAaY;UAEK,6BAA6B;;EAE5C;;EAEA;;EAEA,UAAU,KAAK;;EAEf,aAAa,KAAK;;EAElB,MAAM,KAAK;;EAEX,OAAO,KAAK;;EAEZ,UAAU,KAAK;;EAEf;;EAEA;;EAEA;;EAEA,cAAc;;;;;EAKd;;EAEA;;UAGe,qBAAqB;EACpC;EACA;EACA;EACA,UAAU;EACV,WAAW;EACX;EACA;;UAGe;EACf;EACA;;EAEA;;EAEA;;EAEA;;UAGe;;EAEf;EACA;EACA;EACA;EACA;EACA;;UAGe;;EAEf;;;;;;EAMA;EACA,aAAa;EACb;;EAEA;EACA;EACA;;EAEA;;UAGe;EACf;EACA;;EAEA;;EAEA;;EAEA;;EAEA;EACA,UAAU;;EAEV,WAAW;EACX,UAAU;;UAGK,4BAA4B;EAC3C,cAAc,qBAAqB;EACnC,kBAAkB;EAClB,mBAAmB;;EAEnB,YAAY;;;;;;;;iBAoBE,sBAAsB,MACpC,eAAe,QACf,SAAS,6BAA6B,QACrC,4BAA4B;;;;;;;;;;;;;;;;;;;;;;;;;;;;KChHnB;;UAGK;;EAEf;;EAEA;;EAEA;;;;UAKe;;EAEf;;EAEA;;EAEA;;EAEA;;EAEA,UAAU;;UAGK;EACf;EACA;EACA;EACA;;EAEA;;EAEA;;;EAGA;;EAEA;;EAEA;;UAGe;;EAEf;;EAEA;;EAEA,YAAY;;EAEZ;;EAEA;EACA,MAAM;EACN,OAAO;EACP,SAAS;;EAET;;UAGe;;EAEf;;EAEA;;EAEA;;EAEA;;;EAGA;;UAGe;EACf,SAAS;;EAET;;EAEA;;;;;;;iBAkDc,uBACd,MAAM,iBACN,aAAa,wBACZ;;;;;;;iBAgDa,mBACd,WAAW,iBACX,QAAQ,wBACR,aAAa,wBACZ;;;KA6CS,yBAAyB,uBAAuB,QAAQ;;;UAInD;EACf,QAAQ,QAAQ;EAChB,KAAK,aAAa,eAAe;;;;;;;;cAStB,qBAAqB;;;iBAWlB,wBAAwB,UAAS,eAAoB;;iBAarD,oBAAoB,eAAe;UAyBlC;EACf,QAAQ;EACR,mBAAmB;EACnB,aAAa;;EAEb;;UAGe;EACf;EACA;EACA;EACA;;EAEA,aAAa;;;;;;;;;cAUF;mBACM;mBACA;mBACA;mBACA;EAEjB,YAAY,UAAS;EAOf,OAAO,OAAO,wBAAwB,QAAQ;;;EA6B9C,OACJ,sBACA,KAAK,KAAK;IAAwC;IAAc;MAC/D,QAAQ;EAoBL,IAAI,uBAAuB,QAAQ;EAMnC,QAAQ,QAAQ;;EAKhB,WAAW,uBAAuB,QAAQ"}