TypeScript
Stage 2: wire the goal into runs, add ja/zh, fix metric edge cases
Claude0e4e5c54445c
diff --git a/README.ja.md b/README.ja.mdindex fcaa1f9..f68f5bc 100644--- a/README.ja.md+++ b/README.ja.md@@ -146,7 +146,7 @@ theseus [2m×2a×1h] ▸ /exit| コマンド | 説明 || --- | --- |-| 自由入力 · `/goal <text>` | 達成したい目標(自由入力は目標に蓄積) |+| 自由入力 · `/goal <text>` | 達成したい目標 — 自由入力が蓄積され、run meta に記録されレポートに表示 || `/models` `/agents` `/harness` `/bench` | L0×L1×L2 — 引数なしで一覧、引数ありで選択 || `/runs` `/tasks` `/seed` `/runner` `/judge` `/score` `/lang` | スイープのパラメータ || `/run` · `/view [open]` · `/config` `/clear` `/help` `/exit` | 実行 · 可視化 · 確認/リセット/ヘルプ/終了 |diff --git a/README.ko.md b/README.ko.mdindex 582f5e9..a88ba81 100644--- a/README.ko.md+++ b/README.ko.md@@ -145,7 +145,7 @@ theseus [2m×2a×1h] ▸ /exit| 명령 | 설명 || --- | --- |-| 자유 입력 · `/goal <text>` | 달성하려는 목표(자유 입력은 목표에 누적) |+| 자유 입력 · `/goal <text>` | 달성하려는 목표 — 자유 입력이 누적되고, run meta 에 기록되어 리포트에 표시 || `/models` `/agents` `/harness` `/bench` | L0×L1×L2 — 인자 없으면 목록, 있으면 선택 || `/runs` `/tasks` `/seed` `/runner` `/judge` `/score` `/lang` | 스윕 파라미터 || `/run` · `/view [open]` · `/config` `/clear` `/help` `/exit` | 실행 · 시각화 · 점검/리셋/도움말/종료 |diff --git a/README.md b/README.mdindex 83efbbd..ebec2d7 100644--- a/README.md+++ b/README.md@@ -149,7 +149,7 @@ theseus [2m×2a×1h] ▸ /exit| command | what it does || --- | --- |-| plain text · `/goal <text>` | what you want to achieve (plain text accumulates into the goal) |+| plain text · `/goal <text>` | what you want to achieve — accumulated from plain text, recorded in the run meta, shown on the report || `/models` `/agents` `/harness` `/bench` | the L0×L1×L2 matrix — bare command lists, with args selects || `/runs` `/tasks` `/seed` `/runner` `/judge` `/score` `/lang` | sweep parameters || `/run` | run the comparison with the current selection |diff --git a/README.zh.md b/README.zh.mdindex 7f91297..fabcac9 100644--- a/README.zh.md+++ b/README.zh.md@@ -143,7 +143,7 @@ theseus [2m×2a×1h] ▸ /exit| 命令 | 说明 || --- | --- |-| 自然语言 · `/goal <text>` | 你想达成的目标(自然语言会累加到目标) |+| 自然语言 · `/goal <text>` | 你想达成的目标 — 自然语言累加,记录到 run meta 并显示在报告中 || `/models` `/agents` `/harness` `/bench` | L0×L1×L2 —— 无参数列出,有参数选择 || `/runs` `/tasks` `/seed` `/runner` `/judge` `/score` `/lang` | 扫描参数 || `/run` · `/view [open]` · `/config` `/clear` `/help` `/exit` | 运行 · 可视化 · 查看/重置/帮助/退出 |diff --git a/src/brand.ts b/src/brand.tsindex 200394f..6684ec1 100644--- a/src/brand.ts+++ b/src/brand.ts@@ -27,7 +27,7 @@ export const THEME = {} as const;// Olympus Research 네 원칙(브랜드 보드).-export const PRINCIPLES = {+export const PRINCIPLES: Record<Lang, readonly (readonly [string, string])[]> = {en: [["Alpine Inspiration", "sharp, restrained form — alpine strength and minimalism"],["Focus & Purpose", "the highest peak at the center — exploration toward the summit"],@@ -40,18 +40,34 @@ export const PRINCIPLES = {["테크니컬 엣지", "불필요한 디테일을 덜어내 기능성과 현대적 인상을 강조"],["버서틸리티", "다양한 스케일과 매체에서 명확하게 작동하는 구조"],],-} as const;+ ja: [+ ["アルパイン・インスピレーション", "鋭く抑制された形 — 高山の強靭さとミニマリズム"],+ ["フォーカス&パーパス", "最高峰を中心に — 探究と頂点を目指す精神"],+ ["テクニカル・エッジ", "不要なものを削ぎ落とし、機能性と現代性を強調"],+ ["バーサティリティ", "あらゆるスケール・媒体で明瞭に機能する構造"],+ ],+ zh: [+ ["高山灵感", "锐利而克制的形态 — 高山的坚韧与极简"],+ ["专注与目标", "以最高峰为中心 — 面向顶点的探索精神"],+ ["技术锋芒", "去除多余细节,强调功能性与现代感"],+ ["多场景适应", "在任何尺度与媒介下都清晰可用的结构"],+ ],+};// theseus 의 의미는 "테세우스의 배"(부품을 갈아끼워도 워크플로는 당신의 것). 비주얼은 Olympus 알파인.-export const PHILOSOPHY = {+export const PHILOSOPHY: Record<Lang, string> = {en: "Swap the model, the agent, the harness — plank by plank. The ship is still yours. theseus shows you which planks actually change the voyage.",kr: "모델도, 에이전트도, 하네스도 — 널빤지처럼 갈아끼운다. 그래도 배는 당신의 것. theseus 는 어떤 널빤지가 항해를 실제로 바꾸는지 보여준다.",-} as const;+ ja: "モデルも、エージェントも、ハーネスも — 板を一枚ずつ張り替える。それでも船はあなたのもの。theseus はどの板が航海を本当に変えるかを見せる。",+ zh: "模型、代理、工具链 — 像船板一样逐块更换。船依然是你的。theseus 告诉你哪块船板真正改变了航程。",+};-export const TAGLINE = {+export const TAGLINE: Record<Lang, string> = {en: "compare model × agent × harness on your own tasks",kr: "내 과제로 모델 × 에이전트 × 하네스 비교",-} as const;+ ja: "自分のタスクでモデル × エージェント × ハーネスを比較",+ zh: "用你自己的任务比较模型 × 代理 × 工具链",+};export function philosophy(lang: Lang): string {return PHILOSOPHY[lang] ?? PHILOSOPHY.kr;diff --git a/src/cli.ts b/src/cli.tsindex e227ac5..79917e5 100644--- a/src/cli.ts+++ b/src/cli.ts@@ -54,6 +54,7 @@ function toRunOptions(o: any): RunOptions {language: normalizeLang(o.language),view: normalizeView(o.view),port: Number.parseInt(o.port ?? "7777", 10),+ goal: o.goal || undefined,};}@@ -80,6 +81,7 @@ function addCompareOptions(cmd: Command): Command {.option("--concurrency <n>", "parallel executions", "4").option("--timeout <s>", "per-task timeout seconds", "600").option("--out <dir>", "output dir", "runs")+ .option("--goal <text>", "what you want to achieve (recorded in meta + shown on the report)").option("--judge <name>", "none|heuristic|cc", "none").option("--score", "compute composite score (opt-in)", false).option("--balanced", "difficulty-balanced subset", false)@@ -88,7 +90,7 @@ function addCompareOptions(cmd: Command): Command {.option("--view <mode>", "visualization: inline|open|serve|term", "inline").option("--port <n>", "port for --view serve", "7777").option("--no-logo", "suppress the ASCII logo banner")- .option("-L, --language <lang>", "en|kr", "kr");+ .option("-L, --language <lang>", "en|kr|ja|zh", "kr");}function printPreflight(t: Translator, lang: Lang, opts: RunOptions, yes: boolean): boolean {@@ -235,7 +237,7 @@ function buildProgram(): Command {.aliases(["chat", "repl"]).description("open an interactive session — slash commands to configure & run").option("--out <dir>", "output dir", "runs")- .option("-L, --language <lang>", "en|kr", "kr")+ .option("-L, --language <lang>", "en|kr|ja|zh", "kr").action((o) => runSession({ language: normalizeLang(o.language), outDir: o.out ?? "runs" }));program@@ -243,7 +245,7 @@ function buildProgram(): Command {.description("show recorded results").option("--run <id>", "run id (default: latest)").option("--out <dir>", "output dir", "runs")- .option("-L, --language <lang>", "en|kr", "kr")+ .option("-L, --language <lang>", "en|kr|ja|zh", "kr").action((o) => {const lang = normalizeLang(o.language);const t = translator(lang);@@ -310,7 +312,7 @@ function buildProgram(): Command {.option("--out <dir>", "output dir", "runs").option("--rich", "force cli-table3 fallback", false).option("--open", "open report.html in browser", false)- .option("-L, --language <lang>", "en|kr", "kr")+ .option("-L, --language <lang>", "en|kr|ja|zh", "kr").action((o) =>runTui({out: o.out,@@ -360,7 +362,7 @@ function buildProgram(): Command {.option("--seed <n>", "seed", "0").option("--timeout <s>", "per-task timeout seconds", "120").option("-y, --yes", "allow real runner (not mock)", false)- .option("-L, --language <lang>", "en|kr", "kr")+ .option("-L, --language <lang>", "en|kr|ja|zh", "kr").action(async (o) => {const lang = normalizeLang(o.language);const t = translator(lang);diff --git a/src/i18n.ts b/src/i18n.tsindex 7436c3a..299f8ca 100644--- a/src/i18n.ts+++ b/src/i18n.ts@@ -1,112 +1,243 @@/**- * 다국어 (PRD §16). 사용자에게 보이는 표시 문자열만 i18n. 기본 kr 폴백.+ * 다국어 (PRD §16). 사용자에게 보이는 표시 문자열만 i18n — en/kr/ja/zh (README 4개 언어와 동기).+ * 누락 키 폴백: 요청 언어 → en → kr.*/import type { Lang } from "./models";-type Entry = { en: string; kr: string };+type Entry = { en: string; kr: string; ja?: string; zh?: string };export const STRINGS: Record<string, Entry> = {// CLI / bannerapp_tagline: {en: "theseus — compare model × agent × harness on your own tasks",kr: "theseus — 내 과제로 모델 × 에이전트 × 하네스 비교",+ ja: "theseus — 自分のタスクでモデル × エージェント × ハーネスを比較",+ zh: "theseus — 用你自己的任务比较模型 × 代理 × 工具链",},- preflight_title: { en: "Preflight", kr: "사전점검" },- valid_combos: { en: "Valid combos", kr: "유효 조합" },- excluded_combos: { en: "Excluded combos", kr: "제외 조합" },- scale: { en: "Scale (combos × tasks × runs)", kr: "실행 규모 (조합 × 과제 × 반복)" },- runner_label: { en: "Runner", kr: "러너" },- warnings: { en: "Warnings", kr: "경고" },+ preflight_title: { en: "Preflight", kr: "사전점검", ja: "事前チェック", zh: "预检" },+ valid_combos: { en: "Valid combos", kr: "유효 조합", ja: "有効な組み合わせ", zh: "有效组合" },+ excluded_combos: {+ en: "Excluded combos",+ kr: "제외 조합",+ ja: "除外された組み合わせ",+ zh: "排除的组合",+ },+ scale: {+ en: "Scale (combos × tasks × runs)",+ kr: "실행 규모 (조합 × 과제 × 반복)",+ ja: "実行規模(組み合わせ × タスク × 回数)",+ zh: "执行规模(组合 × 任务 × 次数)",+ },+ runner_label: { en: "Runner", kr: "러너", ja: "ランナー", zh: "运行器" },+ warnings: { en: "Warnings", kr: "경고", ja: "警告", zh: "警告" },gate_blocked: {en: "Blocked: real runner with >30 executions requires --yes",kr: "차단됨: real 러너 30회 초과 실행은 --yes 필요",+ ja: "ブロック: 実ランナーで30回超の実行には --yes が必要",+ zh: "已阻止:真实运行器超过 30 次执行需要 --yes",+ },+ dry_run_only: {+ en: "Dry run — plan only, nothing executed",+ kr: "드라이런 — 계획만, 미실행",+ ja: "ドライラン — 計画のみ、未実行",+ zh: "试运行 — 仅计划,未执行",},- dry_run_only: { en: "Dry run — plan only, nothing executed", kr: "드라이런 — 계획만, 미실행" },- run_complete: { en: "Run complete", kr: "실행 완료" },- report_path: { en: "Report", kr: "리포트" },- graph_path: { en: "Flow graph", kr: "흐름 그래프" },- opening_browser: { en: "opening report in browser…", kr: "브라우저에서 리포트 여는 중…" },- verify_captured: { en: "captured?", kr: "포착됨?" },- run_id: { en: "Run id", kr: "런 ID" },- results_table: { en: "Results", kr: "결과" },+ run_complete: { en: "Run complete", kr: "실행 완료", ja: "実行完了", zh: "运行完成" },+ report_path: { en: "Report", kr: "리포트", ja: "レポート", zh: "报告" },+ graph_path: { en: "Flow graph", kr: "흐름 그래프", ja: "フローグラフ", zh: "流程图" },+ opening_browser: {+ en: "opening report in browser…",+ kr: "브라우저에서 리포트 여는 중…",+ ja: "ブラウザでレポートを開いています…",+ zh: "正在浏览器中打开报告…",+ },+ verify_captured: { en: "captured?", kr: "포착됨?", ja: "捕捉?", zh: "捕获?" },+ run_id: { en: "Run id", kr: "런 ID", ja: "実行ID", zh: "运行 ID" },+ results_table: { en: "Results", kr: "결과", ja: "結果", zh: "结果" },no_models: {en: "No models/agents specified. Example: theseus --model opus4.8 --agent cc --harness none",kr: "모델/에이전트 미지정. 예: theseus --model opus4.8 --agent cc --harness none",+ ja: "モデル/エージェント未指定。例: theseus --model opus4.8 --agent cc --harness none",+ zh: "未指定模型/代理。例:theseus --model opus4.8 --agent cc --harness none",},mock_banner: {en: "MOCK runner — results are synthetic (offline). Not real measurements.",kr: "MOCK 러너 — 합성 결과(오프라인). 실측 아님.",+ ja: "MOCK ランナー — 合成結果(オフライン)。実測ではありません。",+ zh: "MOCK 运行器 — 合成结果(离线)。并非真实测量。",},// dashboard chrome- tab_decide: { en: "Decide", kr: "결정" },- tab_detail: { en: "Detail", kr: "상세" },- tab_flow: { en: "Flow", kr: "흐름" },+ tab_decide: { en: "Decide", kr: "결정", ja: "決定", zh: "决策" },+ tab_detail: { en: "Detail", kr: "상세", ja: "詳細", zh: "详情" },+ tab_flow: { en: "Flow", kr: "흐름", ja: "フロー", zh: "流程" },decide_lead: {en: "The answer at a glance — pick the trade-off, not a single score.",kr: "한눈에 보는 답 — 단일 점수가 아니라 트레이드오프를 고른다.",+ ja: "一目でわかる答え — 単一スコアではなくトレードオフを選ぶ。",+ zh: "一眼看到答案 — 选择权衡,而不是单一分数。",+ },+ sec_parallel: {+ en: "Trade-off across every axis",+ kr: "전 축 트레이드오프",+ ja: "全軸のトレードオフ",+ zh: "全维度权衡",},- sec_parallel: { en: "Trade-off across every axis", kr: "전 축 트레이드오프" },parallel_note: {en: "one line per combo · axes oriented so up = better · hover a line to isolate it",kr: "조합당 한 선 · 위로 갈수록 좋음 · 선에 마우스를 올리면 그 조합만 강조",+ ja: "組み合わせごとに1本 · 上ほど良い · 線にホバーで強調",+ zh: "每个组合一条线 · 越往上越好 · 悬停以突出显示",+ },+ sec_rankings: {+ en: "Ranking on each axis",+ kr: "축별 순위",+ ja: "軸ごとのランキング",+ zh: "各维度排名",},- sec_rankings: { en: "Ranking on each axis", kr: "축별 순위" },- legend_combos: { en: "Combos", kr: "조합" },- axis_better: { en: "better", kr: "좋음" },+ legend_combos: { en: "Combos", kr: "조합", ja: "組み合わせ", zh: "组合" },+ axis_better: { en: "better", kr: "좋음", ja: "良い", zh: "更好" },+ goal_label: { en: "goal", kr: "목표", ja: "目標", zh: "目标" },// report sections- sec_best: { en: "0. Best by perspective", kr: "0. 관점별 최적" },- sec_score: { en: "0.5 Composite score (opt-in)", kr: "0.5 종합점수 (opt-in)" },- sec_accuracy: { en: "1. Accuracy", kr: "1. 정확도" },- sec_efficiency: { en: "2. Efficiency", kr: "2. 효율" },- sec_scatter: { en: "3. Cost-performance scatter", kr: "3. 가성비 산점도" },- pareto_frontier: { en: "Pareto frontier", kr: "파레토 경계" },- philosophy_label: { en: "Ship of Theseus", kr: "테세우스의 배" },- sec_matrix: { en: "4. Task pass/fail matrix", kr: "4. 과제별 통과/실패 매트릭스" },- sec_flow: { en: "5. Subagent flow graphs", kr: "5. 서브에이전트 흐름 그래프" },- sec_failures: { en: "6. Failed tasks & causes", kr: "6. 실패 과제 & 원인" },- sec_uncollected: { en: "7. Uncollected metrics", kr: "7. 미수집 지표" },+ sec_best: {+ en: "0. Best by perspective",+ kr: "0. 관점별 최적",+ ja: "0. 観点別ベスト",+ zh: "0. 各视角最优",+ },+ sec_score: {+ en: "0.5 Composite score (opt-in)",+ kr: "0.5 종합점수 (opt-in)",+ ja: "0.5 総合スコア(opt-in)",+ zh: "0.5 综合评分(可选)",+ },+ sec_accuracy: { en: "1. Accuracy", kr: "1. 정확도", ja: "1. 精度", zh: "1. 准确率" },+ sec_efficiency: { en: "2. Efficiency", kr: "2. 효율", ja: "2. 効率", zh: "2. 效率" },+ sec_scatter: {+ en: "3. Cost-performance scatter",+ kr: "3. 가성비 산점도",+ ja: "3. コスト性能散布図",+ zh: "3. 成本-性能散点图",+ },+ pareto_frontier: {+ en: "Pareto frontier",+ kr: "파레토 경계",+ ja: "パレートフロンティア",+ zh: "帕累托前沿",+ },+ philosophy_label: {+ en: "Ship of Theseus",+ kr: "테세우스의 배",+ ja: "テセウスの船",+ zh: "忒修斯之船",+ },+ sec_matrix: {+ en: "4. Task pass/fail matrix",+ kr: "4. 과제별 통과/실패 매트릭스",+ ja: "4. タスク合否マトリクス",+ zh: "4. 任务通过/失败矩阵",+ },+ sec_flow: {+ en: "5. Subagent flow graphs",+ kr: "5. 서브에이전트 흐름 그래프",+ ja: "5. サブエージェントフローグラフ",+ zh: "5. 子代理流程图",+ },+ sec_failures: {+ en: "6. Failed tasks & causes",+ kr: "6. 실패 과제 & 원인",+ ja: "6. 失敗タスクと原因",+ zh: "6. 失败任务与原因",+ },+ sec_uncollected: {+ en: "7. Uncollected metrics",+ kr: "7. 미수집 지표",+ ja: "7. 未収集の指標",+ zh: "7. 未收集的指标",+ },// perspectives- persp_accuracy: { en: "Highest accuracy", kr: "최고 정확도" },- persp_consistency: { en: "Most consistent", kr: "최고 일관성" },- persp_effcost: { en: "Lowest $/solved", kr: "최저 해결당 비용" },- persp_cost: { en: "Lowest cost", kr: "최저 비용" },- persp_tokens: { en: "Fewest tokens", kr: "최소 토큰" },- persp_time: { en: "Fastest", kr: "최단 시간" },- persp_toolcalls: { en: "Fewest tool calls", kr: "최소 툴콜" },- persp_subagents: { en: "Fewest subagents", kr: "최소 서브에이전트" },- persp_quality: { en: "Highest quality", kr: "최고 품질" },+ persp_accuracy: {+ en: "Highest accuracy",+ kr: "최고 정확도",+ ja: "最高精度",+ zh: "最高准确率",+ },+ persp_consistency: {+ en: "Most consistent",+ kr: "최고 일관성",+ ja: "最も安定",+ zh: "最稳定",+ },+ persp_effcost: {+ en: "Lowest $/solved",+ kr: "최저 해결당 비용",+ ja: "解決あたり最安",+ zh: "每解决最低成本",+ },+ persp_cost: { en: "Lowest cost", kr: "최저 비용", ja: "最低コスト", zh: "最低成本" },+ persp_tokens: { en: "Fewest tokens", kr: "최소 토큰", ja: "最少トークン", zh: "最少 token" },+ persp_time: { en: "Fastest", kr: "최단 시간", ja: "最速", zh: "最快" },+ persp_toolcalls: {+ en: "Fewest tool calls",+ kr: "최소 툴콜",+ ja: "最少ツール呼び出し",+ zh: "最少工具调用",+ },+ persp_subagents: {+ en: "Fewest subagents",+ kr: "최소 서브에이전트",+ ja: "最少サブエージェント",+ zh: "最少子代理",+ },+ persp_quality: { en: "Highest quality", kr: "최고 품질", ja: "最高品質", zh: "最高质量" },// graph- graph_title: { en: "Subagent flow", kr: "서브에이전트 흐름" },- tag_main: { en: "MAIN LOOP", kr: "메인 루프" },- tag_sub: { en: "SUB", kr: "서브" },- node_skills: { en: "Skills", kr: "스킬" },- node_tools: { en: "Tools", kr: "도구" },- node_steps: { en: "Steps", kr: "행동" },- node_input: { en: "Input", kr: "입력" },- node_log: { en: "Log", kr: "로그" },- legend: { en: "Legend", kr: "범례" },- uncollected_note: { en: "null = not collected", kr: "null = 미수집" },+ graph_title: {+ en: "Subagent flow",+ kr: "서브에이전트 흐름",+ ja: "サブエージェントフロー",+ zh: "子代理流程",+ },+ tag_main: { en: "MAIN LOOP", kr: "메인 루프", ja: "メインループ", zh: "主循环" },+ tag_sub: { en: "SUB", kr: "서브", ja: "サブ", zh: "子" },+ node_skills: { en: "Skills", kr: "스킬", ja: "スキル", zh: "技能" },+ node_tools: { en: "Tools", kr: "도구", ja: "ツール", zh: "工具" },+ node_steps: { en: "Steps", kr: "행동", ja: "ステップ", zh: "步骤" },+ node_input: { en: "Input", kr: "입력", ja: "入力", zh: "输入" },+ node_log: { en: "Log", kr: "로그", ja: "ログ", zh: "日志" },+ legend: { en: "Legend", kr: "범례", ja: "凡例", zh: "图例" },+ uncollected_note: {+ en: "null = not collected",+ kr: "null = 미수집",+ ja: "null = 未収集",+ zh: "null = 未收集",+ },// misc columns- col_combo: { en: "combo", kr: "조합" },- col_acc: { en: "accuracy", kr: "정확도" },- col_consistency: { en: "consistency", kr: "일관성" },- col_effcost: { en: "$/solved", kr: "해결당 비용" },+ col_combo: { en: "combo", kr: "조합", ja: "組み合わせ", zh: "组合" },+ col_acc: { en: "accuracy", kr: "정확도", ja: "精度", zh: "准确率" },+ col_consistency: { en: "consistency", kr: "일관성", ja: "安定性", zh: "稳定性" },+ col_effcost: { en: "$/solved", kr: "해결당 비용", ja: "$/解決", zh: "$/解决" },effcost_note: {en: "expected $ to land one passing result, accounting for retries (cost ÷ per-run pass rate)",kr: "통과 결과 1개를 얻기까지의 기대 비용 — 재시도 포함(실행당 비용 ÷ 실행당 통과율)",+ ja: "合格を1つ得るまでの期待コスト — 再試行込み(実行あたりコスト ÷ 実行あたり合格率)",+ zh: "获得一个通过结果的期望成本 — 含重试(每次运行成本 ÷ 每次运行通过率)",+ },+ col_subagents: {+ en: "subagents",+ kr: "서브에이전트",+ ja: "サブエージェント",+ zh: "子代理",},- col_subagents: { en: "subagents", kr: "서브에이전트" },- col_tokens: { en: "total tok", kr: "총 토큰" },- col_cost: { en: "cost", kr: "비용" },- col_wall: { en: "wall", kr: "시간" },- col_toolcalls: { en: "tool calls", kr: "툴콜" },- col_perspective: { en: "perspective", kr: "관점" },- col_value: { en: "value", kr: "값" },+ col_tokens: { en: "total tok", kr: "총 토큰", ja: "総トークン", zh: "总 token" },+ col_cost: { en: "cost", kr: "비용", ja: "コスト", zh: "成本" },+ col_wall: { en: "wall", kr: "시간", ja: "時間", zh: "耗时" },+ col_toolcalls: { en: "tool calls", kr: "툴콜", ja: "ツール呼び出し", zh: "工具调用" },+ col_perspective: { en: "perspective", kr: "관점", ja: "観点", zh: "视角" },+ col_value: { en: "value", kr: "값", ja: "値", zh: "数值" },};export function normalizeLang(input: string | undefined | null): Lang {@@ -114,6 +245,8 @@ export function normalizeLang(input: string | undefined | null): Lang {const s = input.trim().toLowerCase();if (["en", "english", "eng"].includes(s)) return "en";if (["kr", "ko", "korean", "kor", "한국어"].includes(s)) return "kr";+ if (["ja", "jp", "japanese", "jpn", "日本語"].includes(s)) return "ja";+ if (["zh", "cn", "chinese", "zh-cn", "zh-tw", "中文", "汉语", "漢語"].includes(s)) return "zh";return "kr"; // 미상 → 기본 kr}@@ -121,7 +254,7 @@ export function tr(lang: Lang, key: string, fmt?: Record<string, string | numberconst entry = STRINGS[key];let s: string;if (!entry) s = key;- else s = entry[lang] ?? entry.kr ?? key; // 누락 키는 kr 폴백+ else s = entry[lang] ?? entry.en ?? entry.kr ?? key; // 누락 키는 en → kr 폴백if (fmt) {for (const [k, v] of Object.entries(fmt)) {s = s.replaceAll(`{${k}}`, String(v));diff --git a/src/metrics.ts b/src/metrics.tsindex 0327ef8..7b8d994 100644--- a/src/metrics.ts+++ b/src/metrics.ts@@ -73,11 +73,11 @@ function aggregateCost(results: RunResult[]): {} {const actuals = results.map((r) => r.costUsd).filter((c): c is number => c !== null);const ests = results.map((r) => r.estimatedCostUsd).filter((c): c is number => c !== null);- const sources = new Set(results.map((r) => r.costSource));+ // 라벨은 실제 보고되는 값에서 도출 — costUsd:null 인데 costSource:"estimated" 같은 모순 방지.let costSource: CostSource;- if (sources.size === 1 && sources.has("none")) costSource = "none";- else if (sources.has("actual")) costSource = "actual";- else costSource = "estimated";+ if (actuals.length) costSource = "actual";+ else if (ests.length) costSource = "estimated";+ else costSource = "none";return {costUsd: actuals.length ? round4(mean(actuals)) : null,estimatedCostUsd: ests.length ? round4(mean(ests)) : null,diff --git a/src/models.ts b/src/models.tsindex 31243ea..0105790 100644--- a/src/models.ts+++ b/src/models.ts@@ -10,7 +10,7 @@ import { z } from "zod";// 'open' = OpenRouter-hosted models (Qwen, GLM, DeepSeek, Kimi, …; incl. free tier),// usable by model-agnostic agents (opencode/openclaw/hermes).export type Family = "claude" | "gpt" | "open";-export type Lang = "en" | "kr";+export type Lang = "en" | "kr" | "ja" | "zh";export type CostSource = "actual" | "estimated" | "none";export type EventKind = "assistant" | "tool_use" | "subagent" | "result";export type Status = "success" | "fail";@@ -193,6 +193,7 @@ export interface RunOptions {view?: ViewMode; // 시각화 전달 모드 (PRD v2 §W2)port?: number; // serve 모드 포트failRate?: number; // mock 전용 (테스트)+ goal?: string; // 사용자가 밝힌 목표 — meta.json 에 기록되고 report 헤더에 표시}export type ViewMode = "inline" | "open" | "serve" | "term";diff --git a/src/orchestrator.ts b/src/orchestrator.tsindex a280666..31671dd 100644--- a/src/orchestrator.ts+++ b/src/orchestrator.ts@@ -222,6 +222,7 @@ export async function execute(opts: RunOptions): Promise<ExecuteResult> {createdAt: new Date().toISOString(),language: opts.language,mock: anyMock,+ goal: opts.goal,});// report + graph@@ -237,6 +238,7 @@ export async function execute(opts: RunOptions): Promise<ExecuteResult> {judge: judge !== null,score: opts.score,mock: anyMock,+ goal: opts.goal,},recorder.reportPath,);diff --git a/src/preflight.ts b/src/preflight.tsindex 881afe9..b7ee1ea 100644--- a/src/preflight.ts+++ b/src/preflight.ts@@ -1,3 +1,4 @@+import { isLocalBench, taskIds } from "./benches/index";import { isKnownModel, modelFamily } from "./config";import { codexHomeStatus, harnessStatus } from "./harness";/**@@ -60,7 +61,17 @@ export function analyze(}const isReal = opts.runner !== "mock";- const scale = valid.length * Math.max(0, opts.tasks) * Math.max(1, opts.runs);+ // 실제 실행 규모 = 요청 task 수가 아니라 카탈로그로 클램프된 subset 크기(게이트 과대평가 방지).+ let effTasks = Math.max(0, opts.tasks);+ if (isLocalBench(opts.bench)) {+ try {+ effTasks = Math.min(effTasks, taskIds(opts.bench).length);+ } catch {+ // 카탈로그 로드 실패 → 요청값 유지(경고는 실행 시점에)+ }+ }+ if (effTasks === 0) warnings.push("0 tasks selected — the run would be empty");+ const scale = valid.length * effTasks * Math.max(1, opts.runs);const gateBlocked = isReal && scale > REAL_GATE && !opts.yes;return { valid, excluded, warnings, scale, isReal, gateBlocked };diff --git a/src/recorder.ts b/src/recorder.tsindex dd5799a..6a8ef26 100644--- a/src/recorder.ts+++ b/src/recorder.ts@@ -25,15 +25,26 @@ export interface RunMeta {createdAt: string;language: Lang;mock: boolean;+ goal?: string; // 사용자가 세션/CLI 에서 밝힌 목표 — run 의 의도를 기록}+// 밀리초 해상도 충돌 방지 — 같은 ms 에 두 번 만들어지면 -2, -3 … 접미사(런 디렉터리 덮어쓰기 방지).+let lastBaseId = "";+let dupSeq = 1;+export function makeRunId(date = new Date()): string {const p = (n: number, w = 2) => String(n).padStart(w, "0");- return (+ const base =`${date.getUTCFullYear()}${p(date.getUTCMonth() + 1)}${p(date.getUTCDate())}-` +`${p(date.getUTCHours())}${p(date.getUTCMinutes())}${p(date.getUTCSeconds())}-` +- `${p(date.getUTCMilliseconds(), 3)}`- );+ `${p(date.getUTCMilliseconds(), 3)}`;+ if (base === lastBaseId) {+ dupSeq += 1;+ return `${base}-${dupSeq}`;+ }+ lastBaseId = base;+ dupSeq = 1;+ return base;}function safe(s: string): string {diff --git a/src/report.ts b/src/report.tsindex 64d7e81..85ff4fc 100644--- a/src/report.ts+++ b/src/report.ts@@ -26,6 +26,7 @@ export interface ReportInput {judge: boolean;score: boolean;mock: boolean;+ goal?: string;}const COST_MARK = { actual: "✓", estimated: "~", none: "" } as const;@@ -636,6 +637,7 @@ th{background:#1c1f24;color:var(--foam);font-weight:600;letter-spacing:.03em}.hero .meta{display:flex;flex-direction:column;gap:3px}.hero .tag{font-size:13px;color:var(--mist)}.hero .bench{color:var(--foam);font-family:ui-monospace,monospace;font-size:12px}+.hero .goal{color:var(--rope);font-size:12px}/* tabs (CSS-only via radio :checked) */.tabs input{position:absolute;opacity:0;pointer-events:none}.tabnav{display:flex;gap:4px;background:var(--seaDeep);padding:0 24px;border-bottom:1px solid var(--grid)}@@ -686,7 +688,11 @@ th{background:#1c1f24;color:var(--foam);font-weight:600;letter-spacing:.03em}.foot{margin:34px 0 8px;padding-top:14px;border-top:1px solid var(--grid);color:var(--mist);font-size:12px;font-style:italic}${DETAILS_CSS}</style></head><body>-<header class="hero">${svgWordmark()}<div class="meta"><span class="tag">${esc(tr(lang, "app_tagline"))}</span><span class="bench">bench: ${esc(input.bench)}</span></div></header>+<header class="hero">${svgWordmark()}<div class="meta"><span class="tag">${esc(tr(lang, "app_tagline"))}</span><span class="bench">bench: ${esc(input.bench)}</span>${+ input.goal+ ? `<span class="goal">◈ ${esc(tr(lang, "goal_label"))}: ${esc(input.goal)}</span>`+ : ""+ }</div></header><form class="tabs"><input type="radio" name="tab" id="t-decide" checked><input type="radio" name="tab" id="t-detail">diff --git a/src/session.ts b/src/session.tsindex db9302e..b22f075 100644--- a/src/session.ts+++ b/src/session.ts@@ -157,7 +157,7 @@ const HELP = [` ${pc.cyan("/bench")} <name> ${listLocalBenches().join(",")} | terminal-bench-v2`,` ${pc.cyan("/runs")} ${pc.cyan("/tasks")} ${pc.cyan("/seed")} <n> sweep size`,` ${pc.cyan("/runner")} <name> auto|mock|cc|codex|opencode|openclaw|hermes|tb`,- ` ${pc.cyan("/judge")} <name> ${pc.cyan("/score")} on|off ${pc.cyan("/lang")} en|kr`,+ ` ${pc.cyan("/judge")} <name> ${pc.cyan("/score")} on|off ${pc.cyan("/lang")} en|kr|ja|zh`,` ${pc.cyan("/run")} run the comparison → table + report.html`,` ${pc.cyan("/view")} [open] serve every recorded trace at localhost`,` ${pc.cyan("/config")} ${pc.cyan("/clear")} ${pc.cyan("/help")} ${pc.cyan("/exit")}`,@@ -314,6 +314,7 @@ function toRunOptions(s: SessionState): RunOptions {language: s.language,view: "inline",port: s.port,+ goal: s.goal || undefined,};}