niceeval 0.12.0 → 0.12.1-canary.23

package/dist/i18n/en.d.ts

@@ -82,6 +82,25 @@ export declare const en: {
     "cli.accept.flagUnsupported": string;
     "cli.accept.failed": string;
     "cli.accept.done": string;
+    "cli.rename.usage": string;
+    "cli.rename.flagUnsupported": string;
+    "cli.rename.previewHeader": string;
+    "cli.rename.blocked": string;
+    "cli.rename.migratingHeader": string;
+    "cli.rename.migratingRow": string;
+    "cli.rename.excludedHeader": string;
+    "cli.rename.excludedRow": string;
+    "cli.rename.doneHeader": string;
+    "cli.rename.snapshotPath": string;
+    "cli.rename.doneRow": string;
+    "cli.rename.error.sourceEmpty": string;
+    "cli.rename.error.targetNotFound": string;
+    "cli.rename.error.targetHasResults": string;
+    "cli.rename.error.sourceUnreadable": string;
+    "cli.rename.error.artifactUnavailable": string;
+    "cli.rename.error.nothingToMigrate": string;
+    "cli.rename.conflicting": string;
+    "cli.rename.failed": string;
     "cli.error": string;
     "cli.flag.acceptNeedsSelector": string;
     "cli.flag.acceptWithRerunAll": string;

package/dist/i18n/en.js

@@ -79,7 +79,7 @@ export const en = {
     "cli.dry.unit.configs": "configs",
     "cli.dry.affects": "affects {{evals}} · {{ids}}",
     "cli.dry.acceptHint": "accept:  {{command}}",
-    "cli.accept.choiceHeader": "stale  {{selector}}{{change}}  ({{evals}} evals)\n",
+    "cli.accept.choiceHeader": "previous-result  {{selector}}{{change}}  ({{evals}} evals)\n",
     "cli.accept.prompt": "  reuse these results? [y/N] ",
     "cli.accept.nothingToAccept": "No difference in this plan can be accepted (nothing is blocked by the fingerprint gate).\n" +
         "Running as planned.\n",
@@ -91,9 +91,37 @@ export const en = {
         "  fix: pass only @<locator> (and optionally --record <dir>)\n",
     "cli.accept.failed": "error: could not accept result: {{error}}\n",
     "cli.accept.done": "Accepted {{sourceLocator}}. New result locator: {{locator}}. Current fingerprint: {{fingerprint}}\n",
+    "cli.rename.usage": "error: niceeval exp rename expects exactly two arguments: an old id and a new id\n" +
+        "  fix: niceeval exp rename <oldId> <newId> [--dry] [--json]\n",
+    "cli.rename.flagUnsupported": "error: {{flag}} is not valid with niceeval exp rename\n" +
+        "  fix: pass only <oldId> <newId>, optionally --dry / --json\n",
+    "cli.rename.previewHeader": "exp rename preview: {{oldId}} -> {{newId}}\n",
+    "cli.rename.blocked": "  blocked (nothing will be written): {{reason}}\n",
+    "cli.rename.migratingHeader": "  {{count}} terminal results will migrate:\n",
+    "cli.rename.migratingRow": "    {{evalId}}  {{sourceLocator}} -> {{newId}}\n",
+    "cli.rename.excludedHeader": "  {{count}} excluded (not migrated, does not block):\n",
+    "cli.rename.excludedRow": "    {{evalId}}  {{reason}}\n",
+    "cli.rename.doneHeader": "exp rename done: rebound {{count}} terminal results from {{oldId}} to {{newId}}.\n",
+    "cli.rename.snapshotPath": "  new snapshot: {{path}}\n",
+    "cli.rename.doneRow": "    {{evalId}}  {{sourceLocator}} -> {{locator}}\n",
+    "cli.rename.error.sourceEmpty": "error: {{oldId}} has no readable terminal history to migrate to {{newId}}.\n" +
+        "  fix: restore and verify {{oldId}}'s real results before retrying; with no old results, run `niceeval exp {{newId}}` and do not rename.\n" +
+        "       exp rename does not move experiment source, nor delete or rewrite the old result tree.\n",
+    "cli.rename.error.targetNotFound": "error: new id \"{{newId}}\" is not discovered under this project's experiments/.\n" +
+        "  fix: create or rename the experiment in experiments/ first (e.g. `git mv experiments/{{oldId}}.ts experiments/{{newId}}.ts`), then rerun.\n",
+    "cli.rename.error.targetHasResults": "error: {{newId}} already has terminal results for these evals; rename never overwrites existing results.\n" +
+        "  fix: keep the target results, or explicitly clean the target history and re-preview; the command deletes nothing itself.\n",
+    "cli.rename.error.sourceUnreadable": "error: the Record for {{oldId}} is unreadable; cannot migrate to {{newId}}.\n" +
+        "  fix: view this record with a niceeval version that reads its schemaVersion.\n",
+    "cli.rename.error.artifactUnavailable": "error: source evidence cannot be preserved ({{evalId}}); nothing will be written.\n" +
+        "  fix: make the artifact reference and source locator readable, or rerun this eval.\n",
+    "cli.rename.error.nothingToMigrate": "error: nothing to migrate under {{oldId}}: no terminal passed/failed still selected by {{newId}}, or all excluded.\n" +
+        "  fix: check that {{newId}}'s evals selector covers the old experiment's results.\n",
+    "cli.rename.conflicting": "  conflicting evals: {{evals}}\n",
+    "cli.rename.failed": "error: exp rename failed: {{error}}\n",
     "cli.error": "niceeval error: {{error}}\n",
     "cli.flag.acceptNeedsSelector": "error: --accept needs a selector, for example --accept config:judge.model\n" +
-        "  fix: run `niceeval exp <selection> --dry` first; every `stale` line prints the selectors it can accept, copy one verbatim\n" +
+        "  fix: run `niceeval exp <selection> --dry` first; every `previous-result` line prints the selectors it can accept, copy one verbatim\n" +
         "  differences you can accept in this plan: {{available}}\n",
     "cli.flag.acceptWithRerunAll": "--accept cannot be combined with --rerun all: one says trust nothing from cache, the other says trust this difference anyway.\n" +
         "Drop --rerun all to accept the difference, or drop --accept to rerun everything.\n",
@@ -108,8 +136,8 @@ export const en = {
     "runner.setupReturnedCleanup": "{{layer}} returned a function. setup does not carry cleanup and the returned value will not be executed — put the cleanup in the paired teardown of the same layer ({{hint}}); see the experiments tutorial on docs-site or docs/runner.md.\n",
     "runner.experimentTeardownLate": "experiment {{experimentId}}'s teardown was not triggered by the normal countdown path; it has been executed by the end-of-run sweep instead. Record are unaffected; seeing this line means an unlocated intermittent scheduling issue fired — please record this run in the memory ledger.\n",
     "runner.teardownRegistrationWriteFailed": "writing the crash-recovery teardown registration for experiment {{experimentId}} failed: {{message}}. The run continues normally, but a SIGKILL during this run cannot be recovered via `niceeval exp --teardown` or the startup self-heal — check disk space/permissions under .niceeval/teardowns/.\n",
-    "runner.lockTakenOver": "took over an expired case lock for {{experimentId}}/{{evalId}} (previously held by pid {{pid}} on {{host}}; its heartbeat went stale) — that run likely died without releasing it; this run now owns dispatching this case.\n",
-    "runner.gateLeaseTakenOver": "took over an expired concurrency-slot lease for experiment {{experimentId}} (slot {{slot}}, previously held by pid {{pid}} on {{host}}; its heartbeat went stale) — that run likely died without releasing it; this run now owns the slot.\n",
+    "runner.lockTakenOver": "took over an expired case lock for {{experimentId}}/{{evalId}} (previously held by pid {{pid}} on {{host}}; its heartbeat expired) — that run likely died without releasing it; this run now owns dispatching this case.\n",
+    "runner.gateLeaseTakenOver": "took over an expired concurrency-slot lease for experiment {{experimentId}} (slot {{slot}}, previously held by pid {{pid}} on {{host}}; its heartbeat expired) — that run likely died without releasing it; this run now owns the slot.\n",
     "runner.gateLeaseWaiting": "waiting on another run for experiment {{experimentId}}'s concurrency slots: all {{effectiveN}} in use ({{holders}}). Concurrent runs share this experiment's slots, and the smallest maxConcurrency in play wins — this run declared {{declaredN}}. Nothing dispatches until a slot frees up; the other run's slots release when its attempts finish, or 30s after it dies.\n",
     "runner.dispatchHaltedExperiment": "experiment halted (dispatch-halted): {{message}}\n",
     "runner.dispatchHaltedEval": "eval halted: {{message}}\n",
@@ -127,6 +155,8 @@ export const en = {
         "Usage:\n" +
         "  niceeval exp [path|experiment] [eval-id-prefix…]    run experiments\n" +
         "  niceeval exp list [experiment-prefix]               list runnable configs (no dispatch)\n" +
+        "  niceeval exp rename <oldId> <newId>                 rebind terminal results from old to new id\n" +
+        "      --dry   preview without writing; --json   one JSON document\n" +
         "      --teardown   recover a killed run: run only the selected experiments'\n" +
         "        teardown (no attempts, no setup); combining it with eval id prefixes is an error\n" +
         "  niceeval accept @<locator>...                      accept explicit historical results\n" +
@@ -156,8 +186,6 @@ export const en = {
         "      --record <dir>    pin a record root     --exp <id>   repeatable; 2+ compares conditions\n" +
         "      --report <file>   custom report    --page <id>   pick the initial page (multi-page\n" +
         "        reports render it, then list the rest as a page index with copyable commands)\n" +
-        "      --fresh   only count freshly executed attempts (excludes carried-over and\n" +
-        "        historical stitched-in attempts); excluded evals show up as placeholder rows\n" +
         "  niceeval list                                       list discovered evals\n" +
         "  niceeval session list [--all] [experiment-prefix]    query Sessions (read-only)\n" +
         "  niceeval session show <sessionId>                   show one Session (read-only)\n" +
@@ -166,7 +194,6 @@ export const en = {
         "      (same file as show); --page <id> picks the initial page;\n" +
         "      --record <dir> pins a record root; --run <file> opens exactly\n" +
         "      one run; --exp <id> (repeatable) narrows to those experiments;\n" +
-        "      --fresh only new executions\n" +
         "      --out <dir> exports a static site: index.html plus the viewer\n" +
         "      artifacts, ready for any static host\n" +
         "  niceeval sandbox list|enter|history|diff|stop  inspect & destroy sandboxes kept by --keep-sandbox\n" +

package/dist/i18n/zh-CN.d.ts

@@ -73,7 +73,7 @@ export declare const zhCN: {
     readonly "cli.dry.unit.configs": "个运行配置";
     readonly "cli.dry.affects": "影响 {{evals}} · {{ids}}";
     readonly "cli.dry.acceptHint": "accept:  {{command}}";
-    readonly "cli.accept.choiceHeader": "stale  {{selector}}{{change}}  ({{evals}} 条 eval)\n";
+    readonly "cli.accept.choiceHeader": "previous-result  {{selector}}{{change}}  ({{evals}} 条 eval)\n";
     readonly "cli.accept.prompt": "  复用这些结果吗? [y/N] ";
     readonly "cli.accept.nothingToAccept": string;
     readonly "cli.accept.equivalent": "等价命令:  {{command}}\n";
@@ -82,6 +82,25 @@ export declare const zhCN: {
     readonly "cli.accept.flagUnsupported": string;
     readonly "cli.accept.failed": "error: 接受结果失败:{{error}}\n";
     readonly "cli.accept.done": "已接受 {{sourceLocator}}。新结果 locator: {{locator}}。当前指纹:{{fingerprint}}\n";
+    readonly "cli.rename.usage": string;
+    readonly "cli.rename.flagUnsupported": string;
+    readonly "cli.rename.previewHeader": "exp rename 预览:{{oldId}} → {{newId}}\n";
+    readonly "cli.rename.blocked": "  整批阻断,零写入:{{reason}}\n";
+    readonly "cli.rename.migratingHeader": "  将迁移 {{count}} 条终态结果:\n";
+    readonly "cli.rename.migratingRow": "    {{evalId}}  {{sourceLocator}} → {{newId}}\n";
+    readonly "cli.rename.excludedHeader": "  排除 {{count}} 条(不迁移,不阻断):\n";
+    readonly "cli.rename.excludedRow": "    {{evalId}}  {{reason}}\n";
+    readonly "cli.rename.doneHeader": "exp rename 完成:把 {{oldId}} 的 {{count}} 条终态结果重绑到 {{newId}}。\n";
+    readonly "cli.rename.snapshotPath": "  新 snapshot:{{path}}\n";
+    readonly "cli.rename.doneRow": "    {{evalId}}  {{sourceLocator}} → {{locator}}\n";
+    readonly "cli.rename.error.sourceEmpty": string;
+    readonly "cli.rename.error.targetNotFound": string;
+    readonly "cli.rename.error.targetHasResults": string;
+    readonly "cli.rename.error.sourceUnreadable": string;
+    readonly "cli.rename.error.artifactUnavailable": string;
+    readonly "cli.rename.error.nothingToMigrate": string;
+    readonly "cli.rename.conflicting": "  冲突 eval:{{evals}}\n";
+    readonly "cli.rename.failed": "error: exp rename 失败:{{error}}\n";
     readonly "cli.error": "niceeval 出错:{{error}}\n";
     readonly "cli.flag.acceptNeedsSelector": string;
     readonly "cli.flag.acceptWithRerunAll": string;

package/dist/i18n/zh-CN.js

@@ -79,7 +79,7 @@ export const zhCN = {
     "cli.dry.unit.configs": "个运行配置",
     "cli.dry.affects": "影响 {{evals}} · {{ids}}",
     "cli.dry.acceptHint": "accept:  {{command}}",
-    "cli.accept.choiceHeader": "stale  {{selector}}{{change}}  ({{evals}} 条 eval)\n",
+    "cli.accept.choiceHeader": "previous-result  {{selector}}{{change}}  ({{evals}} 条 eval)\n",
     "cli.accept.prompt": "  复用这些结果吗? [y/N] ",
     "cli.accept.nothingToAccept": "本次计划里没有可授权的差异(没有条目卡在指纹门上)。\n" +
         "按原计划运行。\n",
@@ -91,9 +91,37 @@ export const zhCN = {
         "  fix: 只传 @<locator>(可选 `--record <目录>`)\n",
     "cli.accept.failed": "error: 接受结果失败:{{error}}\n",
     "cli.accept.done": "已接受 {{sourceLocator}}。新结果 locator: {{locator}}。当前指纹:{{fingerprint}}\n",
+    "cli.rename.usage": "error: niceeval exp rename 需要恰好两个参数:旧 id 和新 id\n" +
+        "  fix: niceeval exp rename <oldId> <newId> [--dry] [--json]\n",
+    "cli.rename.flagUnsupported": "error: {{flag}} 不能用于 niceeval exp rename\n" +
+        "  fix: 只传 <oldId> <newId>,可选 --dry / --json\n",
+    "cli.rename.previewHeader": "exp rename 预览:{{oldId}} → {{newId}}\n",
+    "cli.rename.blocked": "  整批阻断,零写入:{{reason}}\n",
+    "cli.rename.migratingHeader": "  将迁移 {{count}} 条终态结果:\n",
+    "cli.rename.migratingRow": "    {{evalId}}  {{sourceLocator}} → {{newId}}\n",
+    "cli.rename.excludedHeader": "  排除 {{count}} 条(不迁移,不阻断):\n",
+    "cli.rename.excludedRow": "    {{evalId}}  {{reason}}\n",
+    "cli.rename.doneHeader": "exp rename 完成:把 {{oldId}} 的 {{count}} 条终态结果重绑到 {{newId}}。\n",
+    "cli.rename.snapshotPath": "  新 snapshot:{{path}}\n",
+    "cli.rename.doneRow": "    {{evalId}}  {{sourceLocator}} → {{locator}}\n",
+    "cli.rename.error.sourceEmpty": "error: {{oldId}} 下没有可读的终态历史,无法迁移到 {{newId}}。\n" +
+        "  fix: 恢复并确认 {{oldId}} 的真实结果后重试;没有旧结果时直接运行 `niceeval exp {{newId}}`,不要再执行 rename。\n" +
+        "       exp rename 不移动实验源码,也不删除或改写旧结果树。\n",
+    "cli.rename.error.targetNotFound": "error: 新 id \"{{newId}}\" 没有被当前项目的 experiments/ 发现。\n" +
+        "  fix: 先在 experiments/ 创建或改名出新实验(如 `git mv experiments/{{oldId}}.ts experiments/{{newId}}.ts`),再重跑本命令。\n",
+    "cli.rename.error.targetHasResults": "error: {{newId}} 下已有这些 eval 的终态结果,改名不覆盖已有结果。\n" +
+        "  fix: 保留目标结果继续用,或显式清理目标历史后重新预览;命令自身不删除数据。\n",
+    "cli.rename.error.sourceUnreadable": "error: {{oldId}} 的 Record 读不动,无法迁移到 {{newId}}。\n" +
+        "  fix: 用能读该 schemaVersion 的 niceeval 版本查看这份记录。\n",
+    "cli.rename.error.artifactUnavailable": "error: 来源证据无法保留({{evalId}}),整批零写入。\n" +
+        "  fix: 先确认 artifact 引用与来源 locator 可读,或对这条 eval 重新运行。\n",
+    "cli.rename.error.nothingToMigrate": "error: {{oldId}} 下没有可迁移的 eval:没有 {{newId}} 仍选中的终态 passed/failed,或全部被排除。\n" +
+        "  fix: 检查 {{newId}} 的 evals 选择器是否覆盖旧实验结果。\n",
+    "cli.rename.conflicting": "  冲突 eval:{{evals}}\n",
+    "cli.rename.failed": "error: exp rename 失败:{{error}}\n",
     "cli.error": "niceeval 出错:{{error}}\n",
     "cli.flag.acceptNeedsSelector": "error: --accept 必须带 selector,例如 --accept config:judge.model\n" +
-        "  fix: 先跑 `niceeval exp <选择> --dry`,每条 `stale` 行都打出了它可授权的 selector,原样复制一条\n" +
+        "  fix: 先跑 `niceeval exp <选择> --dry`,每条 `previous-result` 行都打出了它可授权的 selector,原样复制一条\n" +
         "  本次计划里可以授权的差异:{{available}}\n",
     "cli.flag.acceptWithRerunAll": "--accept 不能与 --rerun all 同用:一边全不采信缓存,一边又要采信这条差异,方向自相矛盾。\n" +
         "要授权这条差异就去掉 --rerun all;要全量重跑就去掉 --accept。\n",
@@ -127,6 +155,8 @@ export const zhCN = {
         "用法:\n" +
         "  niceeval exp [路径|实验] [eval-id 前缀…]   跑实验\n" +
         "  niceeval exp list [实验前缀]              列出可运行的实验配置(不派发)\n" +
+        "  niceeval exp rename <旧 id> <新 id>        把旧实验的终态结果重绑到新 id(显式审计迁移)\n" +
+        "      --dry   只预览不写盘;--json   输出单份 JSON 文档\n" +
         "      --teardown   强杀后补收尾:只对选中的实验各执行一次 teardown(不派发\n" +
         "        attempt、不跑 setup);与 eval id 前缀组合是用法错误\n" +
         "  niceeval accept @<locator>...           接受明确列出的历史结果\n" +
@@ -153,8 +183,6 @@ export const zhCN = {
         "      --record <目录>  钉死记录根   --exp <id> 可重复,两个以上进入对照\n" +
         "      --report <文件> 自定义报告   --page <id> 定初始页(多页报告渲染该页,\n" +
         "        尾部再附其余页索引)\n" +
-        "      --fresh   只统计新执行的 attempt(排除携带条目与跨快照拼入的历史执行);\n" +
-        "        被排除的题转成占位行\n" +
         "  niceeval list                            列出发现到的 eval\n" +
         "  niceeval session list [--all] [实验前缀]  查询 Session(只读)\n" +
         "  niceeval session show <sessionId>         查看一个 Session(只读)\n" +
@@ -162,7 +190,6 @@ export const zhCN = {
         "      报告页 + 证据室;--report <文件> 整槽换成自定义报告(与 show 同一文件)\n" +
         "      --page <id> 定初始页   --record <目录> 钉死记录根\n" +
         "      --run <文件> 只打开这一份快照   --exp <id>(可重复)收窄到这些实验;\n" +
-        "      --fresh 只看新执行\n" +
         "      --out <目录> 静态导出:index.html 连同查看器 artifact,可直接静态托管\n" +
         "  niceeval sandbox list|enter|history|diff|stop  查看与销毁 --keep-sandbox 留下的现场\n" +
         "  niceeval sandbox list --orphans / prune         核对并收回被强杀留下的无主实例\n" +

package/dist/record/open.js

@@ -30,8 +30,8 @@ const locatorIndexByResults = new WeakMap();
 export class LocatorNotFoundError extends Error {
     locator;
     constructor(locator) {
-        super(`No attempt found for locator "${locator}" in this results root. It may be stale ` +
-            "(the run was deleted, or publish didn't include it) or mistyped.");
+        super(`No attempt found for locator "${locator}" in this results root. It may refer to an older record ` +
+            "(the run was deleted, or publish didn't include it) or be mistyped.");
         this.locator = locator;
         this.name = "LocatorNotFoundError";
     }

package/dist/record/types.d.ts

@@ -1,4 +1,4 @@
-import type { DiagnosticRecord, EvalResult, ExperimentRunInfo, LocalizedText, SandboxBuildRecord, TimingActivity } from "../types.ts";
+import type { DiagnosticRecord, EvalResult, ExperimentRunInfo, LocalizedText, SandboxBuildRecord, TimingActivity, Verdict } from "../types.ts";
 import type { CommandExitEvidence, O11ySummary, StreamEvent, TraceSpan } from "../types.ts";
 import type { AgentSetupManifest, DiffData, SourceArtifact } from "../types.ts";
 import type { AttemptIdentity, AttemptLocator } from "./locator.ts";
@@ -204,7 +204,7 @@ export interface Record {
 }
 /**
  * 一个实验的覆盖事实:已知 eval 并集(分母)与当前口径下没有任何 attempt 的题。
- * `missingEvalIds` 永远被算出来,不静默——渲染面把它转成覆盖占位行
+ * `missing` 永远被算出来,不静默——渲染面把它转成覆盖占位行
  * (见 docs/feature/sample/library.md「选择快照」「时效:新执行与历史执行」)。
  *
  * `run` 是该 Experiment 的锚点 Run:零 attempt 的 Eval 按 agent / model / flags 归组时
@@ -217,8 +217,20 @@ export interface SampleCoverage {
     run: Run;
     /** 分母:本地历史 ∪ 各快照携带的 knownEvalIds,交命令行范围(与 `exp.knownEvalIds` 同源)。 */
     knownEvalIds: string[];
-    /** 当前口径下没有任何 attempt 的题(含 `fresh: true` 排除历史执行后新产生的缺口)。 */
-    missingEvalIds: string[];
+    /** 当前配置下没有物理 Attempt 的题，以及帮助用户决定下一步的缺口原因。 */
+    missing: SampleMissing[];
+}
+/** 当前结果缺口；原因只解释下一步，不构成另一种结果状态。 */
+export interface SampleMissing {
+    evalId: string;
+    /** 从未有物理 Attempt，或有历史结果但没有一条能代表当前配置。 */
+    reason: "never-run" | "previous-result";
+    /** 最近一条旧结果的审计入口；它不参与当前统计，也不保证一定满足 accept 资格。 */
+    previous?: {
+        locator: AttemptLocator;
+        verdict: Verdict;
+        startedAt: string;
+    };
 }
 /**
  * Sample(范围):选出的快照 + 口径 + 已按口径物化的 attempt 全集 + 挑选警告。
@@ -228,8 +240,6 @@ export interface SampleCoverage {
 export interface Sample {
     /** 这份 Sample 的口径,字面写在数据上。 */
     mode: "latest-run" | "current";
-    /** 是否只保留各 Experiment 锚点中真实执行的 attempt。 */
-    fresh: boolean;
     runs: Run[];
     /**
      * 按口径物化的 attempt 全集:消费 attempts 就自动正确,不需要自己 flatten runs,
@@ -251,7 +261,6 @@ export interface Sample {
         evals?: string | string[];
     }): Sample;
     filter(predicate: (attempt: AttemptHandle) => boolean): Sample;
-    freshOnly(): Sample;
 }
 /**
  * 挑选警告:每种带 kind、可判断的结构化字段和渲染好的英文 message;能用一条命令直接推进的

package/dist/report/built-in/result-components.js

@@ -1,13 +1,11 @@
 import { jsx as _jsx, jsxs as _jsxs } from "react/jsx-runtime";
 import { defineComponent } from "../definition/tree.js";
-import { Callouts, Chart, Col, Conversation, CopyBlock, DiffView, Grid, Scatter, Series, SourceView, Stat, TableContentView, Text, Waterfall, } from "../definition/primitives.js";
+import { Callouts, Chart, Col, Conversation, CopyBlock, DiffView, Grid, Scatter, Series, SourceView, Stat, TableContentView, Waterfall, } from "../definition/primitives.js";
 import { HeroCard } from "../components/site-components/index.js";
 import { AttemptSummary } from "../components/attempt-detail/index.js";
 import { experimentListContent } from "../components/entity-lists/content.js";
-import { ExperimentTableView } from "../components/entity-lists/index.js";
 import { attemptAssertionsContent, attemptConversationContent, attemptDiffContent, embedConversationInSource, executionEvidenceUnavailableCallouts, attemptFixPromptContent, attemptNoticesContent, projectedSourceContent, attemptTimelineContent, } from "../components/attempt-detail/content.js";
 import { stabilityMatrixContent } from "../slices/content.js";
-import { formatInstant, formatReportDateTimeRange, } from "../model/format.js";
 import { DEFAULT_REPORT_LOCALE, localeText } from "../model/locale.js";
 function experimentTarget(point) {
     return { page: "experiment", params: { experiment: point.key } };
@@ -15,27 +13,15 @@ function experimentTarget(point) {
 function summaryView(result, locale) {
     const snapshot = result.summary;
     const tally = 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", 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, "scopeSummary.votesEval"), value: { kind: "verdict", counts: tally } }), _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] }));
+    return (_jsx(Col, { className: "niceeval-sample-summary", 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, "scopeSummary.votesEval"), value: { kind: "verdict", counts: tally } }), _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 })] }) }));
 }
 export const StandardOverviewResultView = defineComponent(async ({ result }, ctx) => {
     const table = experimentListContent(result.experiments);
-    const freshTable = result.freshExperiments ? experimentListContent(result.freshExperiments) : null;
     const hasPassRate = table.columns.some((column) => column.key === "passRate");
     const hasTotalScore = table.columns.some((column) => column.key === "totalScore");
     const defaultSort = hasPassRate === hasTotalScore
@@ -43,7 +29,7 @@ export const StandardOverviewResultView = defineComponent(async ({ result }, ctx
         : hasPassRate
             ? "passRate"
             : "totalScore";
-    return (_jsxs(Col, { children: [_jsx(HeroCard, { title: ctx.report.title, data: result.hero }), _jsx(Callouts, { items: result.notices }), _jsx(Callouts, { items: result.diagnostics }), _jsx(CopyBlock, { content: result.fixPrompt }), summaryView(result, DEFAULT_REPORT_LOCALE), _jsx(Col, { children: result.charts.map((chart) => chart.y === "passRate" ? (_jsx(Scatter, { points: chart.points, x: "costUSD", y: "passRate", point: "experiment", series: "series", connect: chart.connect, pointTarget: experimentTarget, legend: true }, chart.y)) : (_jsx(Scatter, { points: chart.points, x: "costUSD", y: "totalScore", point: "experiment", series: "series", connect: chart.connect, pointTarget: experimentTarget, legend: true }, chart.y))) }), freshTable === null ? (_jsx(TableContentView, { data: table, sort: defaultSort, searchable: true })) : (_jsx(ExperimentTableView, { fullContent: table, freshContent: freshTable, sort: defaultSort, searchable: true }))] }));
+    return (_jsxs(Col, { children: [_jsx(HeroCard, { title: ctx.report.title, data: result.hero }), _jsx(Callouts, { items: result.notices }), _jsx(Callouts, { items: result.diagnostics }), _jsx(CopyBlock, { content: result.fixPrompt }), summaryView(result, DEFAULT_REPORT_LOCALE), _jsx(Col, { children: result.charts.map((chart) => chart.y === "passRate" ? (_jsx(Scatter, { points: chart.points, x: "costUSD", y: "passRate", point: "experiment", series: "series", connect: chart.connect, pointTarget: experimentTarget, legend: true }, chart.y)) : (_jsx(Scatter, { points: chart.points, x: "costUSD", y: "totalScore", point: "experiment", series: "series", connect: chart.connect, pointTarget: experimentTarget, legend: true }, chart.y))) }), _jsx(TableContentView, { data: table, sort: defaultSort, searchable: true })] }));
 });
 StandardOverviewResultView.displayName = "StandardOverviewResultView";
 function stabilityPresentation(result) {

package/dist/report/components/entity-lists/compute.d.ts

@@ -11,12 +11,6 @@ export declare function failureSummaryOf(result: EvalResult): {
     summary: string | null;
     more: number;
 };
-/**
- * 「只看新执行」开关在场的判据(docs/feature/reports/components/summaries/experiment-table.md
- * 「只看新执行」):Sample 里既没有历史执行也没有过期结论时不画开关——一个永远不改变行集的
- * 控件只会让人怀疑自己看漏了什么。`ExperimentTable` 与内建默认报告共用同一条判据。
- */
-export declare function hasHistoricalOrStale(items: readonly ExperimentListItem[]): boolean;
 /** 已选出的 AttemptHandle[] → 列表行；顺序保持传入顺序（不再次按 Sample 去重）。 */
 export declare function attemptRowsOf(attempts: readonly AttemptHandle[]): Promise<AttemptListItem[]>;
 /** `attemptListData(input)`:每个 Attempt 一项,顺序取自 Sample 展平顺序(不重排)。 */

package/dist/report/components/entity-lists/compute.js

@@ -8,11 +8,11 @@
 // - core 中立:只认 Metric / Dimension 接口,不出现具体 agent 名的分支。
 import { comparabilityConfigOf, deepEqualJson } from "../../../sample/index.js";
 import { foldEvalVerdict } from "../../../shared/verdict.js";
-import { collectItems, computeCell, evalIdOf, experimentIdOf, fullEvalKey, groupItems, historicalOf, locatorOf, msSince, resolveInput, staleReferenceOf, } from "../../model/aggregate.js";
+import { collectItems, computeCell, evalIdOf, experimentIdOf, fullEvalKey, groupItems, locatorOf, resolveInput, } from "../../model/aggregate.js";
 import { attemptCostUSD, costUSD, durationMs, examScore, passRate, tokens, totalScore } from "../../model/metrics.js";
 import { compactAssertionSummary, primaryAssertionSummary, summaryText } from "../../../assertions/display.js";
 import { firstLine } from "../../../util.js";
-import { selectedAttemptsOnly, summarizeItems } from "../shared-compute.js";
+import { summarizeItems } from "../shared-compute.js";
 /**
  * 一次 attempt 的单行结果摘要(断言摘要契约):failed 取主失败断言摘要(不含
  * "+N more",N 单独进 moreFailures),errored 取结构化 error 的一层摘要
@@ -47,11 +47,6 @@ export function failureSummaryOf(result) {
 async function attemptListItemOf(item) {
     const result = item.attempt.result;
     const { summary, more } = failureSummaryOf(result);
-    const historical = historicalOf(item);
-    // 缺 startedAt(legacy / 第三方落盘)时退化到所属快照的 startedAt——时效标注宁可粗一档
-    // 时距,不留空字段(与 dedupeAttempts「缺才不去重」同一条「不伪造」纪律,这里伪造的只是
-    // 展示粒度,不影响身份判定)。
-    const startedAt = result.startedAt ?? item.run.startedAt;
     return {
         experimentId: experimentIdOf(item),
         evalId: evalIdOf(item),
@@ -63,23 +58,13 @@ async function attemptListItemOf(item) {
         moreFailures: more,
         examScore: await computeCell(examScore, [item]),
         totalScore: await computeCell(totalScore, [item]),
+        tokens: await computeCell(tokens, [item]),
         durationMs: result.durationMs,
         costUSD: attemptCostUSD(result),
-        startedAt,
-        historical,
-        ...(historical ? { staleSinceMs: msSince(startedAt) } : {}),
+        startedAt: result.startedAt ?? item.run.startedAt,
         locator: locatorOf(item),
     };
 }
-/**
- * 「只看新执行」开关在场的判据(docs/feature/reports/components/summaries/experiment-table.md
- * 「只看新执行」):Sample 里既没有历史执行也没有过期结论时不画开关——一个永远不改变行集的
- * 控件只会让人怀疑自己看漏了什么。`ExperimentTable` 与内建默认报告共用同一条判据。
- */
-export function hasHistoricalOrStale(items) {
-    return items.some((item) => Object.keys(item.staleReferences).length > 0 ||
-        item.evalRows.some((row) => row.attempts.some((attempt) => attempt.historical)));
-}
 /** 已选出的 AttemptHandle[] → 列表行；顺序保持传入顺序（不再次按 Sample 去重）。 */
 export async function attemptRowsOf(attempts) {
     return Promise.all(attempts.map((attempt) => attemptListItemOf({ attempt, run: attempt.run, watermark: attempt.run })));
@@ -170,38 +155,6 @@ function byMetricDescThenId(valueOf) {
         return vb - va || a.experimentId.localeCompare(b.experimentId);
     };
 }
-/**
- * 覆盖缺口两档占位的「过期结论」参考(docs/feature/reports/components/summaries/experiment-table.md
- * 「覆盖缺口的两档占位行」):`missingEvalIds` 里、`historyAttempts` 中存在与 `anchorConfigHash`
- * 不可比判定的题,取其中最近一条。`fresh` 为 true 时整份不给参考——读者已声明只看新执行,
- * 占位行就不再把被排除的历史结论请回来。
- */
-function staleReferencesFor(experimentId, missingEvalIds, historyAttempts, anchorConfigHash, fresh) {
-    if (fresh || missingEvalIds.length === 0)
-        return {};
-    const missing = new Set(missingEvalIds);
-    const candidatesByEval = new Map();
-    for (const attempt of historyAttempts) {
-        if (attempt.experimentId !== experimentId)
-            continue;
-        if (!missing.has(attempt.evalId))
-            continue;
-        if (attempt.run.configHash === anchorConfigHash)
-            continue; // 可比,不是「过期结论」候选
-        const list = candidatesByEval.get(attempt.evalId);
-        if (list)
-            list.push(attempt);
-        else
-            candidatesByEval.set(attempt.evalId, [attempt]);
-    }
-    const out = {};
-    for (const [evalId, candidates] of candidatesByEval) {
-        const reference = staleReferenceOf(candidates);
-        if (reference)
-            out[evalId] = reference;
-    }
-    return out;
-}
 /**
  * `experimentListData(input)`:每个 experiment 一项,展开到每道 Eval;初始排序按这份列表
  * 自身的题型构成选择主读数——纯通过制沿用端到端通过率降序,纯计分制改按总分降序(缺数据
@@ -212,7 +165,7 @@ function staleReferencesFor(experimentId, missingEvalIds, historyAttempts, ancho
  * 看跨配置演化用 run 维度或 MetricLine,不把两套配置拼成一行冒充单一配置。
  */
 export async function experimentListData(input) {
-    const { runs, attempts, coverage, historyAttempts, fresh } = resolveInput(input);
+    const { runs, attempts, coverage } = resolveInput(input);
     const coverageByExperiment = new Map(coverage.map((c) => [c.experimentId, c]));
     // 可比性配置单义检查:同一 experiment 的输入快照必须共享一套可比性配置。
     const configByExperiment = new Map();
@@ -229,7 +182,7 @@ export async function experimentListData(input) {
                 'use the "run" dimension or MetricLine; to show the current level, pass results.current() which selects a single config per experiment.');
         }
     }
-    const items = collectItems(runs, selectedAttemptsOnly(attempts));
+    const items = collectItems(runs, attempts);
     const groups = groupItems(items, "experiment");
     const out = [];
     for (const [experimentId, group] of groups) {
@@ -260,7 +213,7 @@ export async function experimentListData(input) {
         }
         const experiment = newest.run.experiment ?? newest.attempt.result.experiment;
         const model = newest.attempt.result.model ?? newest.run.model;
-        const missingEvalIds = coverageByExperiment.get(experimentId)?.missingEvalIds ?? [];
+        const coverageEntry = coverageByExperiment.get(experimentId);
         out.push({
             experimentId,
             agent: newest.run.agent || newest.attempt.result.agent,
@@ -275,15 +228,14 @@ export async function experimentListData(input) {
             tokens: await computeCell(tokens, group),
             evals: stats.evals,
             attempts: stats.attempts,
-            historicalAttempts: group.filter(historicalOf).length,
-            missingEvalIds,
-            staleReferences: staleReferencesFor(experimentId, missingEvalIds, historyAttempts, coverageByExperiment.get(experimentId)?.run.configHash, fresh),
+            knownEvalIds: coverageEntry?.knownEvalIds ?? [],
+            missing: coverageEntry?.missing ?? [],
             lastRunAt: stats.lastRunAt,
             evalRows,
         });
     }
-    // coverage 不是 attempt 的附属品：--fresh 和 current() 都可能让一个实验当前口径下
-    // 零 attempt。仍然给它一行，让 missingEvalIds 的占位题可达，不能把整实验静默吞掉。
+    // coverage 不是 attempt 的附属品:current() 可能让一个实验当前口径下零 attempt。仍然给它
+    // 一行,让 missing 的占位题可达,不能把整实验静默吞掉。
     for (const coverageEntry of coverage) {
         if (groups.has(coverageEntry.experimentId))
             continue;
@@ -307,9 +259,8 @@ export async function experimentListData(input) {
             tokens: await computeCell(tokens, emptyItems),
             evals: 0,
             attempts: 0,
-            historicalAttempts: 0,
-            missingEvalIds: coverageEntry.missingEvalIds,
-            staleReferences: staleReferencesFor(coverageEntry.experimentId, coverageEntry.missingEvalIds, historyAttempts, anchor.configHash, fresh),
+            knownEvalIds: coverageEntry.knownEvalIds,
+            missing: coverageEntry.missing,
             lastRunAt: anchor.startedAt,
             evalRows: [],
         });

package/dist/report/components/entity-lists/content.d.ts

@@ -1,7 +1,5 @@
 import type { TableContent } from "../../definition/cell.ts";
 import type { AttemptListItem, EvalListItem, ExperimentListItem } from "../../model/types.ts";
-/** 覆盖构成副行的 key 前缀;测试与消费方靠它把这一行从 Eval / 组行里筛出去。 */
-export declare const COVERAGE_ROW_PREFIX = "coverage:";
 export declare function experimentListContent(items: readonly ExperimentListItem[]): TableContent;
 export declare function evalListContent(items: readonly EvalListItem[]): TableContent;
 export declare function attemptListContent(items: readonly AttemptListItem[]): TableContent;