niceeval 0.12.1 → 0.12.2-canary.31

package/dist/report/components/site-components/projections.d.ts

@@ -1,12 +0,0 @@
-import type { ReportInput } from "../../model/types.ts";
-import type { CalloutGroup } from "../../definition/primitives/callouts-logic.ts";
-import type { CopyBlockContent } from "../../definition/primitives/copy-block.tsx";
-import type { WaterfallContent } from "../../definition/primitives/waterfall.tsx";
-/** SampleIssue → CalloutGroup[]（文案在 Source 侧收口；Callouts 只渲染）。 */
-export declare function sampleNoticesContent(input: ReportInput): Promise<readonly CalloutGroup[]>;
-/** Run diagnostics → CalloutGroup[]。 */
-export declare function runNoticesContent(input: ReportInput): Promise<readonly CalloutGroup[]>;
-/** Fix prompt → CopyBlockContent | null。 */
-export declare function sampleFixPromptContent(input: ReportInput): Promise<CopyBlockContent | null>;
-/** TraceWaterfallRow[] → WaterfallContent。 */
-export declare function sampleTracesContent(input: ReportInput): Promise<WaterfallContent>;

package/dist/report/components/site-components/projections.js

@@ -1,67 +0,0 @@
-// 专用件 → 原语 Content 的投影。Source compute 产出原语可直接吃的形状。
-import { groupScopeWarnings } from "./scope-warnings.js";
-import { copyFixPromptData, scopeWarningsData, snapshotDiagnosticsData, traceWaterfallData } from "./compute.js";
-/** SampleIssue → CalloutGroup[]（文案在 Source 侧收口；Callouts 只渲染）。 */
-export async function sampleNoticesContent(input) {
-    const issues = await scopeWarningsData(input);
-    if (issues.length === 0)
-        return [];
-    // Content 携带已解析字符串(LocalizedText 的 string 形态);宿主 locale 切换不改写已烘文案。
-    const { groups } = groupScopeWarnings(issues, "en");
-    return groups.map((group) => ({
-        title: group.title,
-        ...(group.headCommand !== null ? { command: group.headCommand } : {}),
-        badges: group.badges.map((badge) => badge.text),
-        items: group.issues.map((issue) => ({
-            level: issue.code === "unreadable-run" ? "error" : "warning",
-            message: issue.detail,
-            ...(issue.action !== null ? { command: issue.action } : {}),
-        })),
-    }));
-}
-/** Run diagnostics → CalloutGroup[]。 */
-export async function runNoticesContent(input) {
-    const rows = await snapshotDiagnosticsData(input);
-    if (rows.length === 0)
-        return [];
-    return rows.map((row) => ({
-        title: `${row.experimentId} · ${row.startedAt}`,
-        items: row.diagnostics.map((d) => ({
-            level: d.level === "error" ? "error" : "warning",
-            message: d.detail,
-        })),
-    }));
-}
-/** Fix prompt → CopyBlockContent | null。 */
-export async function sampleFixPromptContent(input) {
-    const data = await copyFixPromptData(input);
-    if (data.failures === 0 || !data.prompt)
-        return null;
-    return {
-        title: `Fix prompt · ${data.failures} failure${data.failures === 1 ? "" : "s"}`,
-        text: data.prompt,
-    };
-}
-/** TraceWaterfallRow[] → WaterfallContent。 */
-export async function sampleTracesContent(input) {
-    const rows = await traceWaterfallData(input);
-    return rows.map((row) => {
-        const nodes = row.spans.map((span, i) => ({
-            key: `${row.locator}:${i}`,
-            label: span.name,
-            kind: span.kind,
-            startOffsetMs: span.startOffsetMs,
-            durationMs: span.durationMs,
-            ...(span.failed ? { failed: true } : {}),
-        }));
-        return {
-            key: row.locator,
-            // label 是 experimentId/evalId;locator 由行头单独呈现,label 不重复它
-            // (docs/feature/reports/components/primitives/waterfall.md)。
-            label: `${row.experimentId}/${row.evalId}`,
-            durationMs: row.durationMs,
-            nodes,
-            locator: row.locator,
-        };
-    });
-}

package/dist/report/components/site-components/scope-warnings.d.ts

@@ -1,44 +0,0 @@
-import type { SampleIssue } from "../../../record/types.ts";
-import { type ReportLocale } from "../../model/locale.ts";
-export interface SampleIssueNotice {
-    code: string;
-    title: string;
-    detail: string;
-    action: string | null;
-    experimentId?: string;
-    issue: SampleIssue;
-}
-/** Reports 层的 Issue→Notice policy；文案和动作不写回 Sample。 */
-export declare const NoticeCatalog: {
-    of(issue: SampleIssue, locale: ReportLocale): SampleIssueNotice;
-};
-export interface ScopeWarningGroup {
-    /** 实验组为 experimentId;kind 组为登记的组头文案(含条数);未登记 kind 用 kind 原文。 */
-    title: string;
-    /** 每条警告一枚、与 issues 同序;未登记徽标模板的成员不出徽标。 */
-    badges: readonly {
-        kind: string;
-        text: string;
-    }[];
-    /** 组内命令去重后恰一条时归组头(复制即推进整组);多条或零条为 null,命令随明细逐条走。 */
-    headCommand: string | null;
-    /** 原始条目(明细层,message 单源)。 */
-    issues: readonly SampleIssueNotice[];
-}
-export interface GroupedScopeWarnings {
-    /** 分类计数汇总行,任何组数下都产出;web 面用作外层折叠块的 <summary>,text 面只在多组时打印。 */
-    summary: string;
-    groups: readonly ScopeWarningGroup[];
-    /** 警告总条数 ≤ 3 时组级明细默认展开(web 面第二层 <details> 的 open;阈值是行为契约,无开关)。 */
-    detailsOpen: boolean;
-}
-/** 明细折叠块的标签(「N 条原始警告」)。 */
-export declare function warningDetailsLabel(locale: ReportLocale, n: number): string;
-/**
- * 按「用户要做什么」组织,不按发生顺序:带 experimentId 且登记了徽标模板的 kind 按实验聚合,
- * 其余(含未登记 kind)按 kind 聚合。组排序:实验作用域组在前(按实验 id 字典序),
- * 非实验作用域组在后(按 kind);类别两档制(integrity / freshness)已随 stale-run /
- * partial-coverage 一并删除——三种 warning kind 都是同一类完整性事实,不再需要档位区分
- * (docs/feature/reports/components/summaries/sample-notices.md「聚合轴是动作,不是发生顺序」)。
- */
-export declare function groupScopeWarnings(input: readonly SampleIssue[], locale: ReportLocale): GroupedScopeWarnings;

package/dist/report/components/site-components/scope-warnings.js

@@ -1,115 +0,0 @@
-// ScopeWarnings 的聚合层:先把 Sample Issue 解释为 Notice,再按「下一步动作」组织成组,
-// web / text 两面共用(docs/feature/reports/components/summaries/sample-notices.md「聚合轴是动作,不是发生顺序」)。
-import { localeText } from "../../model/locale.js";
-/** Reports 层的 Issue→Notice policy；文案和动作不写回 Sample。 */
-export const NoticeCatalog = {
-    of(issue, locale) {
-        switch (issue.code) {
-            case "unfinished-run":
-                return {
-                    code: issue.code,
-                    title: localeText(locale, "issues.badge.unfinishedSnapshot"),
-                    detail: `${issue.experimentId} has an unfinished run from ${issue.startedAt}.`,
-                    action: `niceeval exp ${issue.experimentId}`,
-                    experimentId: issue.experimentId,
-                    issue,
-                };
-            case "dangling-evidence":
-                return {
-                    code: issue.code,
-                    title: "unavailable evidence",
-                    detail: `${issue.experimentId}/${issue.evalId} attempt ${issue.attempt} has unavailable evidence.`,
-                    action: null,
-                    experimentId: issue.experimentId,
-                    issue,
-                };
-            case "unreadable-run":
-                return {
-                    code: issue.code,
-                    title: pluralText(locale, "issues.group.unreadableSnapshot", 1),
-                    detail: `${issue.dir} is unreadable (${issue.reason}).`,
-                    action: null,
-                    issue,
-                };
-        }
-    },
-};
-/** 实验作用域且登记了徽标模板的 kind 才进实验组;其余(含未登记 kind)按 kind 聚合。 */
-const EXPERIMENT_KINDS = new Set(["unfinished-run"]);
-function pluralText(locale, base, n) {
-    return localeText(locale, `${base}.${n === 1 ? "one" : "other"}`, { n });
-}
-/** 明细折叠块的标签(「N 条原始警告」)。 */
-export function warningDetailsLabel(locale, n) {
-    return pluralText(locale, "issues.details", n);
-}
-function badgeText(w, locale) {
-    switch (w.code) {
-        case "unfinished-run":
-            return localeText(locale, "issues.badge.unfinishedSnapshot");
-        default:
-            return null;
-    }
-}
-/** 组内命令去重:恰一条时它就是「复制即推进整组」的组头命令。 */
-function dedupeCommand(members) {
-    const actions = [...new Set(members.map((member) => member.action).filter((action) => action !== null))];
-    return actions.length === 1 ? actions[0] : null;
-}
-/**
- * 按「用户要做什么」组织,不按发生顺序:带 experimentId 且登记了徽标模板的 kind 按实验聚合,
- * 其余(含未登记 kind)按 kind 聚合。组排序:实验作用域组在前(按实验 id 字典序),
- * 非实验作用域组在后(按 kind);类别两档制(integrity / freshness)已随 stale-run /
- * partial-coverage 一并删除——三种 warning kind 都是同一类完整性事实,不再需要档位区分
- * (docs/feature/reports/components/summaries/sample-notices.md「聚合轴是动作,不是发生顺序」)。
- */
-export function groupScopeWarnings(input, locale) {
-    const issues = input.map((issue) => NoticeCatalog.of(issue, locale));
-    const byExperiment = new Map();
-    const byKind = new Map();
-    for (const w of issues) {
-        if (EXPERIMENT_KINDS.has(w.code) && typeof w.experimentId === "string") {
-            const members = byExperiment.get(w.experimentId) ?? [];
-            members.push(w);
-            byExperiment.set(w.experimentId, members);
-        }
-        else {
-            const members = byKind.get(w.code) ?? [];
-            members.push(w);
-            byKind.set(w.code, members);
-        }
-    }
-    const groups = [];
-    // 实验作用域组在前,按实验 id 字典序(Map 插入顺序取决于扫描顺序,这里显式排序)。
-    for (const experimentId of [...byExperiment.keys()].sort()) {
-        const members = byExperiment.get(experimentId);
-        groups.push({
-            title: experimentId,
-            badges: members
-                .map((w) => ({ kind: w.code, text: badgeText(w, locale) }))
-                .filter((b) => b.text !== null),
-            headCommand: dedupeCommand(members),
-            issues: members,
-        });
-    }
-    // 非实验作用域组在后,按 kind 字典序。
-    for (const kind of [...byKind.keys()].sort()) {
-        const members = byKind.get(kind);
-        groups.push({
-            title: kind === "unreadable-run" ? pluralText(locale, "issues.group.unreadableSnapshot", members.length) : kind,
-            badges: [],
-            headCommand: dedupeCommand(members),
-            issues: members,
-        });
-    }
-    const parts = [];
-    if (byExperiment.size > 0)
-        parts.push(pluralText(locale, "issues.summary.experiments", byExperiment.size));
-    for (const kind of [...byKind.keys()].sort()) {
-        const members = byKind.get(kind);
-        parts.push(kind === "unreadable-run"
-            ? pluralText(locale, "issues.group.unreadableSnapshot", members.length)
-            : `${kind} ×${members.length}`);
-    }
-    return { summary: parts.join(" · "), groups, detailsOpen: issues.length <= 3 };
-}

package/dist/report/components/site-components/snapshot-diagnostics.d.ts

@@ -1,18 +0,0 @@
-import type { DiagnosticRecord } from "../../../types.ts";
-import type { SnapshotDiagnosticsData, SnapshotDiagnosticsItem } from "../../model/types.ts";
-import { type ReportLocale } from "../../model/locale.ts";
-export interface SnapshotDiagnosticsGroup {
-    experimentId: string;
-    /** 该实验贡献的来源快照,已按 startedAt 新到旧排列;各自的 diagnostics 不跨快照合并。 */
-    items: readonly SnapshotDiagnosticsItem[];
-}
-export interface GroupedSnapshotDiagnostics {
-    /** 汇总行:涉及多少个 experiment、多少个 Run、多少条记录(按 count 计数)与最高严重度。 */
-    summary: string;
-    groups: readonly SnapshotDiagnosticsGroup[];
-    /** 全部记录里的最高严重度;没有 error 记录时为 warning。空集合时也是 warning(不参与渲染)。 */
-    severity: "warning" | "error";
-}
-/** 一条 DiagnosticRecord 折叠后的出现次数;省略等于 1。 */
-export declare function occurrencesOf(record: DiagnosticRecord): number;
-export declare function groupSnapshotDiagnostics(data: SnapshotDiagnosticsData, locale: ReportLocale): GroupedSnapshotDiagnostics;

package/dist/report/components/site-components/snapshot-diagnostics.js

@@ -1,42 +0,0 @@
-// SnapshotDiagnostics 的聚合层:把 snapshotDiagnosticsData 的产物按来源 experiment 分组,
-// web / text 两面共用(docs/feature/reports/library.md)。
-// 输入已经由 snapshotDiagnosticsData 按 experiment id 字典序、组内 startedAt 新到旧排好序,
-// 这里只按 experimentId 分桶,不重新排序、不跨来源合并 DiagnosticRecord。
-import { localeText } from "../../model/locale.js";
-function pluralText(locale, base, n) {
-    return localeText(locale, `${base}.${n === 1 ? "one" : "other"}`, { n });
-}
-/** 一条 DiagnosticRecord 折叠后的出现次数;省略等于 1。 */
-export function occurrencesOf(record) {
-    return record.count ?? 1;
-}
-export function groupSnapshotDiagnostics(data, locale) {
-    const byExperiment = new Map();
-    for (const item of data) {
-        const members = byExperiment.get(item.experimentId) ?? [];
-        members.push(item);
-        byExperiment.set(item.experimentId, members);
-    }
-    const groups = [...byExperiment.entries()].map(([experimentId, items]) => ({
-        experimentId,
-        items,
-    }));
-    let records = 0;
-    let severity = "warning";
-    for (const item of data) {
-        for (const d of item.diagnostics) {
-            records += occurrencesOf(d);
-            if (d.level === "error")
-                severity = "error";
-        }
-    }
-    const summary = data.length === 0
-        ? ""
-        : [
-            pluralText(locale, "snapshotDiagnostics.summary.experiments", byExperiment.size),
-            pluralText(locale, "snapshotDiagnostics.summary.runs", data.length),
-            pluralText(locale, "snapshotDiagnostics.summary.records", records),
-            localeText(locale, severity === "error" ? "snapshotDiagnostics.severity.error" : "snapshotDiagnostics.severity.warning"),
-        ].join(" · ");
-    return { summary, groups, severity };
-}

package/dist/report/components/site-components/validators.d.ts

@@ -1,6 +0,0 @@
-import { type Validator } from "../shared.ts";
-export declare const validateHeroData: Validator;
-export declare const validateScopeWarningsData: Validator;
-export declare const validateSnapshotDiagnosticsData: Validator;
-export declare const validateCopyFixPromptData: Validator;
-export declare const validateTraceWaterfallData: Validator;

package/dist/report/components/site-components/validators.js

@@ -1,115 +0,0 @@
-// site-components 的 *Data 结构校验(与旧专用件 data 形态同契约,供 validate.test 守护)。
-import { arrayProblem, isObject } from "../shared.js";
-export const validateHeroData = (data) => {
-    if (!isObject(data))
-        return "expected an object";
-    if (!(data.latestStartedAt === null || typeof data.latestStartedAt === "string")) {
-        return '"latestStartedAt" must be a string or null';
-    }
-    if (typeof data.runs !== "number")
-        return '"runs" must be a number';
-    return null;
-};
-const UNREADABLE_SNAPSHOT_REASONS = ["incompatible", "malformed", "incomplete"];
-function scopeWarningProblem(value, path) {
-    if (!isObject(value))
-        return `"${path}" must be a SampleIssue object`;
-    if (typeof value.code !== "string")
-        return `"${path}.code" must be a string`;
-    switch (value.code) {
-        case "unfinished-run":
-            if (typeof value.experimentId !== "string")
-                return `"${path}.experimentId" must be a string`;
-            if (typeof value.startedAt !== "string")
-                return `"${path}.startedAt" must be a string`;
-            if (typeof value.dir !== "string")
-                return `"${path}.dir" must be a string`;
-            return null;
-        case "dangling-evidence":
-            if (typeof value.experimentId !== "string")
-                return `"${path}.experimentId" must be a string`;
-            if (typeof value.evalId !== "string")
-                return `"${path}.evalId" must be a string`;
-            if (typeof value.attempt !== "number")
-                return `"${path}.attempt" must be a number`;
-            if (typeof value.artifactBase !== "string")
-                return `"${path}.artifactBase" must be a string`;
-            if (!Array.isArray(value.artifacts) || !value.artifacts.every((artifact) => typeof artifact === "string")) {
-                return `"${path}.artifacts" must be an array of strings`;
-            }
-            return null;
-        case "unreadable-run":
-            if (typeof value.dir !== "string")
-                return `"${path}.dir" must be a string`;
-            if (typeof value.reason !== "string" || !UNREADABLE_SNAPSHOT_REASONS.includes(value.reason)) {
-                return `"${path}.reason" must be one of ${JSON.stringify(UNREADABLE_SNAPSHOT_REASONS)}`;
-            }
-            return null;
-        default:
-            return `"${path}.code" is not a supported SampleIssue code`;
-    }
-}
-export const validateScopeWarningsData = (data) => arrayProblem(data, "data", scopeWarningProblem);
-function diagnosticRecordProblem(value, path) {
-    if (!isObject(value))
-        return `"${path}" must be a DiagnosticRecord { code, level, message, phase }`;
-    if (typeof value.code !== "string")
-        return `"${path}.code" must be a string`;
-    if (value.level !== "warning" && value.level !== "error")
-        return `"${path}.level" must be "warning" or "error"`;
-    if (typeof value.message !== "string")
-        return `"${path}.message" must be a string`;
-    if (typeof value.phase !== "string")
-        return `"${path}.phase" must be a string`;
-    return null;
-}
-function snapshotDiagnosticsItemProblem(value, path) {
-    if (!isObject(value))
-        return `"${path}" must be a SnapshotDiagnosticsItem { experimentId, startedAt, diagnostics }`;
-    if (typeof value.experimentId !== "string")
-        return `"${path}.experimentId" must be a string`;
-    if (typeof value.startedAt !== "string")
-        return `"${path}.startedAt" must be a string`;
-    return arrayProblem(value.diagnostics, `${path}.diagnostics`, diagnosticRecordProblem);
-}
-export const validateSnapshotDiagnosticsData = (data) => arrayProblem(data, "data", snapshotDiagnosticsItemProblem);
-export const validateCopyFixPromptData = (data) => {
-    if (!isObject(data))
-        return "expected an object";
-    if (typeof data.prompt !== "string")
-        return '"prompt" must be a string';
-    if (typeof data.failures !== "number")
-        return '"failures" must be a number';
-    return null;
-};
-const TRACE_SPAN_KINDS = ["agent", "model", "tool", "other"];
-function traceSpanSummaryProblem(value, path) {
-    if (!isObject(value))
-        return `"${path}" must be a TraceSpanSummary { name, kind, startOffsetMs, durationMs, failed }`;
-    if (typeof value.name !== "string")
-        return `"${path}.name" must be a string`;
-    if (typeof value.kind !== "string" || !TRACE_SPAN_KINDS.includes(value.kind)) {
-        return `"${path}.kind" must be one of ${JSON.stringify(TRACE_SPAN_KINDS)}`;
-    }
-    if (typeof value.startOffsetMs !== "number")
-        return `"${path}.startOffsetMs" must be a number`;
-    if (typeof value.durationMs !== "number")
-        return `"${path}.durationMs" must be a number`;
-    if (typeof value.failed !== "boolean")
-        return `"${path}.failed" must be a boolean`;
-    return null;
-}
-export const validateTraceWaterfallData = (data) => arrayProblem(data, "data", (row, path) => {
-    if (!isObject(row))
-        return `"${path}" must be an object`;
-    if (typeof row.experimentId !== "string")
-        return `"${path}.experimentId" must be a string`;
-    if (typeof row.evalId !== "string")
-        return `"${path}.evalId" must be a string`;
-    if (typeof row.locator !== "string")
-        return `"${path}.locator" must be a string`;
-    if (!(row.durationMs === null || typeof row.durationMs === "number")) {
-        return `"${path}.durationMs" must be a number or null`;
-    }
-    return arrayProblem(row.spans, `${path}.spans`, traceSpanSummaryProblem);
-});

package/dist/report/components/summaries/compute.d.ts

@@ -1,7 +0,0 @@
-import type { ReportInput, SampleSummaryContent } from "../../model/types.ts";
-/**
- * `sampleSummary(input)`:范围摘要——快照时间窗、experiment / eval / attempt 数、
- * 两级判定计票、端到端通过率与总成本(docs/feature/reports/components/summaries/sample-summary.md)。
- * data 恒携带两级计票;通过率来自官方两级指标引擎,不从任一计票重算。
- */
-export declare function sampleSummary(input: ReportInput): Promise<SampleSummaryContent>;

package/dist/report/components/summaries/compute.js

@@ -1,60 +0,0 @@
-// 计算函数(*Data):ReportInput → 一份组件数据。SampleSummary 的 sampleSummary 住在这里
-// (docs/feature/reports/components/summaries/README.md)。SampleOverview 是纯组合组件,不产生
-// 独立的 data,没有对应的计算函数。
-//
-// 共同约定(docs/feature/reports/architecture.md「指标聚合不变量」):
-// - 第一参收 ReportInput = Sample | readonly Run[];issues 不进组件数据(宿主统一显示);
-// - 聚合前按身份键去重(dedupeAttempts;missing-startedAt 不去重、如实保留、不透出警告);
-// - null ≠ 0:缺数据不编数,覆盖率经 samples/total 如实暴露。
-import { collectItems, computeCell, resolveInput } from "../../model/aggregate.js";
-import { costUSD, passRate, totalScore } from "../../model/metrics.js";
-import { evaluationKindComposition } from "../../model/evaluation-kind.js";
-import { selectedAttemptsOnly, summarizeItems, tallyOf } from "../shared-compute.js";
-// ───────────────────────── sampleSummary ─────────────────────────
-/** costUSD 的求和投影:两级都 sum(题内多轮求和 + 跨题求和 = 全量求和)。 */
-const totalCostMetric = {
-    name: "total-cost",
-    label: costUSD.label,
-    unit: "$",
-    value: costUSD.value,
-    perEval: "sum",
-    acrossEvals: "sum",
-};
-/**
- * `sampleSummary(input)`:范围摘要——快照时间窗、experiment / eval / attempt 数、
- * 两级判定计票、端到端通过率与总成本(docs/feature/reports/components/summaries/sample-summary.md)。
- * data 恒携带两级计票;通过率来自官方两级指标引擎,不从任一计票重算。
- */
-export async function sampleSummary(input) {
-    const { runs, attempts: scopeAttempts } = resolveInput(input);
-    const items = collectItems(runs, selectedAttemptsOnly(scopeAttempts));
-    let earliest = null;
-    let latest = null;
-    for (const run of runs) {
-        if (earliest === null || run.startedAt < earliest)
-            earliest = run.startedAt;
-        if (latest === null || run.startedAt > latest)
-            latest = run.startedAt;
-    }
-    const stats = summarizeItems(items);
-    const attemptVerdicts = tallyOf();
-    for (const item of items)
-        attemptVerdicts[item.attempt.result.verdict] += 1;
-    // 题型构成:决定渲染面的主 KPI 是通过率、总分,还是两者都显示。单点判据见
-    // evaluationKindComposition()(docs/feature/reports/library/measures.md「题型构成与主读数」)——
-    // 不在这里另设一份 hasPoints/hasPass 判断。
-    const composition = await evaluationKindComposition(input);
-    const passItems = items.filter((item) => item.attempt.result.evaluationKind !== "points");
-    return {
-        range: { earliestStartedAt: earliest, latestStartedAt: latest },
-        experiments: stats.experiments,
-        evals: stats.evals,
-        attempts: stats.attempts,
-        evalVerdicts: stats.verdicts,
-        attemptVerdicts,
-        endToEndPassRate: await computeCell(passRate, passItems),
-        evaluationKindComposition: composition,
-        ...(composition !== "pass" ? { totalScore: await computeCell(totalScore, items) } : {}),
-        totalCostUSD: await computeCell(totalCostMetric, items),
-    };
-}

package/dist/report/components/summaries/index.d.ts

@@ -1,29 +0,0 @@
-import type { Sample } from "../../../record/types.ts";
-import type { ChartTargetPoint } from "../../definition/primitives/chart.tsx";
-import type { ReportTarget } from "../../definition/report.ts";
-import type { SeriesInput } from "../../model/types.ts";
-import type { ChromeProps } from "../shared.ts";
-export { validateSampleSummaryContent } from "./validate.ts";
-export { StabilityOverview, type StabilityOverviewProps } from "./stability-overview.tsx";
-export type SampleSummaryProps = ChromeProps & {
-    input?: Sample;
-    votes?: "eval" | "attempt";
-};
-export declare const SampleSummary: import("../../definition/tree.ts").ReportComponent<SampleSummaryProps>;
-type ComparisonChrome = ChromeProps & {
-    connect?: boolean;
-};
-export type ExperimentScatterProps = ComparisonChrome & {
-    input?: Sample;
-    series?: SeriesInput;
-    /**
-     * 点该指向谁(components/summaries/experiment-scatter.md「默认点目标」)。省略时落到
-     * `defaultExperimentPointTarget`——点身份恒为 experiment id(`point="experiment"`),
-     * 目标是 experiment 详情页;报告没有声明该 id 的页时 `ctx.href` 自然给不出链接,
-     * 点退化成纯图形,不是这里判断"页存不存在"。
-     */
-    pointTarget?: (point: ChartTargetPoint) => ReportTarget | undefined;
-};
-export type SampleOverviewProps = ExperimentScatterProps;
-export declare const ExperimentScatter: import("../../definition/tree.ts").ReportComponent<ExperimentScatterProps>;
-export declare const SampleOverview: import("../../definition/tree.ts").ReportComponent<ExperimentScatterProps>;

package/dist/report/components/summaries/index.js

@@ -1,122 +0,0 @@
-import { jsx as _jsx, jsxs as _jsxs } from "react/jsx-runtime";
-import { defineComponent } from "../../definition/tree.js";
-import { Col, Grid, Scatter, Stat, Text } from "../../definition/primitives.js";
-import { resolveInput, seriesName } from "../../model/aggregate.js";
-import { label } from "../../model/flag.js";
-import { agent, aggregate, costUSD, experiment, model, passRate, totalScore, } from "../../model/calculation.js";
-import { evaluationKindComposition } from "../../model/evaluation-kind.js";
-import { DEFAULT_REPORT_LOCALE, localeText } from "../../model/locale.js";
-import { formatInstant, formatReportDateTimeRange } from "../../model/format.js";
-import { toSummaryItems } from "../../model/conversions.js";
-import { ExperimentTable } from "../entity-lists/index.js";
-export { validateSampleSummaryContent } from "./validate.js";
-export { StabilityOverview } from "./stability-overview.js";
-export const SampleSummary = defineComponent(async (props, ctx) => {
-    const input = props.input ?? ctx.scope;
-    const snapshot = await toSummaryItems(input);
-    const locale = props.locale ?? DEFAULT_REPORT_LOCALE;
-    const votes = props.votes ?? "eval";
-    const tally = votes === "attempt" ? snapshot.attemptVerdicts : snapshot.evalVerdicts;
-    const formattedRange = snapshot.range.earliestStartedAt !== null && snapshot.range.latestStartedAt !== null
-        ? formatReportDateTimeRange(snapshot.range.earliestStartedAt, snapshot.range.latestStartedAt, locale)
-        : null;
-    return (_jsxs(Col, { className: ["niceeval-sample-summary", props.className].filter(Boolean).join(" "), children: [_jsxs(Grid, { children: [snapshot.evaluationKindComposition !== "points" ? (_jsx(Stat, { label: localeText(locale, "scopeSummary.passRate"), value: { kind: "metric", metric: snapshot.endToEndPassRate } })) : null, snapshot.totalScore !== undefined ? (_jsx(Stat, { label: localeText(locale, "scopeSummary.totalScore"), value: { kind: "metric", metric: snapshot.totalScore } })) : null, _jsx(Stat, { label: localeText(locale, "scopeSummary.experiments"), value: snapshot.experiments }), _jsx(Stat, { label: localeText(locale, "scopeSummary.evals"), value: snapshot.evals }), _jsx(Stat, { label: localeText(locale, "scopeSummary.attempts"), value: snapshot.attempts }), _jsx(Stat, { label: localeText(locale, votes === "attempt" ? "scopeSummary.votesAttempt" : "scopeSummary.votesEval"), value: {
-                            kind: "verdict",
-                            counts: {
-                                passed: tally.passed,
-                                failed: tally.failed,
-                                errored: tally.errored,
-                                skipped: tally.skipped,
-                            },
-                        } }), _jsx(Stat, { label: localeText(locale, "scopeSummary.totalCost"), value: { kind: "metric", metric: snapshot.totalCostUSD }, detail: snapshot.totalCostUSD.samples < snapshot.totalCostUSD.total
-                            ? localeText(locale, "scopeSummary.costCoverage", {
-                                samples: snapshot.totalCostUSD.samples,
-                                total: snapshot.totalCostUSD.total,
-                            })
-                            : undefined })] }), snapshot.range.latestStartedAt !== null ? (_jsx(Text, { className: "niceeval-sample-summary-range", children: snapshot.range.earliestStartedAt !== null && snapshot.range.earliestStartedAt !== snapshot.range.latestStartedAt
-                    ? localeText(locale, "scopeSummary.runRange", {
-                        from: formattedRange.from,
-                        to: formattedRange.to,
-                    })
-                    : localeText(locale, "scopeSummary.lastRun", {
-                        time: formatInstant(snapshot.range.latestStartedAt, locale),
-                    }) })) : null] }));
-});
-SampleSummary.displayName = "SampleSummary";
-/**
- * `ExperimentScatter` 的默认点目标:点身份键固定是 experiment id(`scatterBlock` 恒传
- * `point="experiment"`),`Chart` 内核把它原样落成该点的 Dataset 行 key——不需要反查
- * 证据即可拿到这个点是哪个 experiment。
- */
-function defaultExperimentPointTarget(point) {
-    return { page: "experiment", params: { experiment: point.key } };
-}
-const LINE_LABEL_KEY = "line";
-function resolveComparisonSeries(input, props) {
-    const hasLine = resolveInput(input).runs.some((s) => s.experiment?.labels?.[LINE_LABEL_KEY] !== undefined);
-    const series = props.series ?? (hasLine ? label(LINE_LABEL_KEY) : "agent");
-    return { series, connect: props.connect ?? seriesName(series) === LINE_LABEL_KEY };
-}
-function filterInputByEvaluationKind(input, evaluationKind) {
-    return input.filter((attempt) => (attempt.result.evaluationKind === "points" ? "points" : "pass") === evaluationKind);
-}
-function seriesGroup(series) {
-    if (typeof series === "string") {
-        if (series === "agent")
-            return { key: "agent", fn: agent };
-        if (series === "model")
-            return { key: "model", fn: model };
-        if (series === "experiment")
-            return { key: "experiment", fn: experiment };
-        throw new Error(`SampleOverview series "${series}" is not a built-in group — use "agent" | "model" | "experiment", or label()/flag().`);
-    }
-    if (Array.isArray(series)) {
-        throw new Error("SampleOverview does not accept composite series arrays; pass a single dimension.");
-    }
-    if ("kind" in series && (series.kind === "label" || series.kind === "flag")) {
-        const name = series.name;
-        const kind = series.kind;
-        const fn = (subject) => {
-            const bag = kind === "label" ? subject.run.experiment?.labels : subject.run.experiment?.flags;
-            const value = bag?.[name];
-            return value === undefined || value === null ? "(missing)" : String(value);
-        };
-        Object.defineProperty(fn, "name", { value: `${kind}:${name}` });
-        return { key: name, fn };
-    }
-    if ("of" in series && typeof series.of === "function") {
-        throw new Error(`SampleOverview cannot use CustomDimension "${series.name}" with aggregate() — CustomDimension reads AttemptHandle; write a GroupFunction over AggregationSubject instead.`);
-    }
-    throw new Error(`SampleOverview series kind "${series.kind ?? "unknown"}" is not supported for aggregate grouping yet.`);
-}
-async function scatterBlock(input, options) {
-    const group = seriesGroup(options.series);
-    const yCalc = options.y === "totalScore" ? totalScore : passRate;
-    const points = await aggregate(input, {
-        by: {
-            experiment,
-            [group.key]: group.fn,
-        },
-        values: {
-            costUSD,
-            [options.y]: yCalc,
-        },
-    });
-    return (_jsx(Scatter, { points: points, x: "costUSD", y: options.y, point: "experiment", series: group.key, connect: options.connect, pointTarget: options.pointTarget, locale: options.locale, legend: true }));
-}
-export const ExperimentScatter = defineComponent(async (props, ctx) => {
-    const input = props.input ?? ctx.scope;
-    const { series, connect } = resolveComparisonSeries(input, props);
-    const composition = await evaluationKindComposition(input);
-    const pointTarget = props.pointTarget ?? defaultExperimentPointTarget;
-    if (composition !== "mixed") {
-        const primary = composition === "points" ? "totalScore" : "passRate";
-        return (_jsx(Col, { className: props.className, children: await scatterBlock(input, { series, connect, y: primary, pointTarget, locale: props.locale }) }));
-    }
-    const passInput = filterInputByEvaluationKind(input, "pass");
-    const pointsInput = filterInputByEvaluationKind(input, "points");
-    return (_jsxs(Col, { className: props.className, children: [await scatterBlock(passInput, { series, connect, y: "passRate", pointTarget, locale: props.locale }), await scatterBlock(pointsInput, { series, connect, y: "totalScore", pointTarget, locale: props.locale })] }));
-});
-ExperimentScatter.displayName = "ExperimentScatter";
-export const SampleOverview = defineComponent((props) => (_jsxs(Col, { className: props.className, children: [_jsx(SampleSummary, { input: props.input, locale: props.locale }), _jsx(ExperimentScatter, { input: props.input, series: props.series, connect: props.connect, pointTarget: props.pointTarget, locale: props.locale }), _jsx(ExperimentTable, { input: props.input, locale: props.locale })] })));
-SampleOverview.displayName = "SampleOverview";