Repository navigation

feat(demo): rebuild the demo as a single-page workbench with the OSW v2.1 baseline and an official-leaderboard comparison - #7

Merged

xxxyxun merged 31 commits into

AI45Lab:mainfrom

two-tiger:feature/demo-rebuild

Oct 27, 2026

Merged

feat(demo): rebuild the demo as a single-page workbench with the OSW v2.1 baseline and an official-leaderboard comparison#7
xxxyxun merged 31 commits into
AI45Lab:mainfrom
two-tiger:feature/demo-rebuild

Commits on Oct 26, 2026
- feat(tools): add baseline summary, report fill, run compare and rank probe
  root
  committed
  5189a2f
- fix(tools): count every decision without an executable action as a no-op
  root
  committed
  1e44a08
- feat(demo): 流式读取层 demo/reader（题号归一 + 大文件保护）
  tandem-demo
  committed
  ae7108b
- feat(demo): 聚合层委托 tools/batch_summary + Sai 逐题对照列
  tandem-demo
  committed
  6e66cb3
- feat(demo): 官方榜/Sai 逐题分快照 + 价格表 + 刷新脚本
  tandem-demo
  committed
  53a7fe3
- docs(demo): 落地重建计划（stage+memory 批次 × 官方 harness 对照）
  tandem-demo
  committed
  78adf8b
- feat(demo): 对照汇总（compare_summary）+ 满分浮点容差
  tandem-demo
  committed
  aa86436
- feat(demo): 单页工作台骨架 + 总览视图（重写 demo 前端）
  tandem-demo
  committed
  e4b774a
- feat(demo): 官方对照视图 + 机制视图
  tandem-demo
  committed
  e71a977
- feat(demo): 轨迹页 + 双轨同步播放器
  tandem-demo
  committed
  6b97585
- feat(demo): 发布闸门——导出到内网共享目录 + 诚实披露红线自检
  tandem-demo
  committed
  9fa6cd6
- feat(demo): 重做首页视觉 + 官方 Sai 双边轨迹对照
  tandem-demo
  committed
  cedf2ad
- fix(demo): 对照页左侧加载我们的逐题数据（原来传了 null）
  tandem-demo
  committed
  d38423e
- docs(demo): 补实施后记（计划偏差与实施期发现的两个真 bug）
  tandem-demo
  committed
  75ea77b
- feat(demo): 官方参考轨迹补帧映射（12 题 · 2298 张帧）
  tandem-demo
  committed
  5e185bf
- feat(demo): 修首页崩溃 + 首页简化放大 + 双播放器 + 机制泳道
  tandem-demo
  committed
  e1a4ca3
- fix(demo): 逐步归因对齐到「一次观察 = 一帧 = 一次动作」
  tandem-demo
  committed
  4a4018b
- feat(demo): 首页按反馈精简
  tandem-demo
  committed
  ce2e0ac
- feat(demo): 「官方对照」改为「轨迹浏览」，按 OSW2 showcase 布局重做三栏
  tandem-demo
  committed
  8e905fd
- fix(demo): Tandem 轨迹无图 + 整宽排版 + 读数移到播放器下
  tandem-demo
  committed
  8538906
- feat(demo): 首页重排为「简介 → 机制 → 结果」滚动页
  tandem-demo
  committed
  75429a5
- fix(demo): 修首页「view is not a function」+ 加渲染冒烟测试
  tandem-demo
  committed
  238dab7
- test(demo): 渲染冒烟的 JS stub 拆成独立文件（避开 ruff 行长）
  tandem-demo
  committed
  9d80308
- fix(demo): 机制段左侧吸顶列滚下去就空了 + 规划原文取不到
  tandem-demo
  committed
  6533374
Commits on Oct 27, 2026
- feat(demo): 接入 OSW v2.1 全量基线（c10+c20 合并 108 题）
  tandem-demo
  committed
  9a13590
- feat(demo): 机制段改成可交互流程图（不再是纯文字）
  tandem-demo
  committed
  fb81e46
- fix(demo): 修「首页没有内联的总览数据」+ 构建脚本加 --site-only 快路径
  tandem-demo
  committed
  12d68fb
- feat(demo): 首页按子任务循环重画机制段（流程图+粘性卡片+真实case泳道）
  
  tandem-demo
  and
  MiniMax-M3.1-Flash
  committed
  545762b
- refactor(demo): 拆前端超长模块（browser.js / views.css）
  
  tandem-demo
  and
  MiniMax-M3.1-Flash
  committed
  79f109c
- fix(demo): 修流程图出框/压字 + 补真实子任务与动作级候选 + 删真实case泳道
  
  tandem-demo
  and
  MiniMax-M3.1-Flash
  committed
  e6044a6
- docs(demo): 记下机制段流程图的视觉 smoke 链路（本机无浏览器时的替代方案）
  
  tandem-demo
  and
  MiniMax-M3.1-Flash
  committed
  5e6e4af

Commits

Commits on Oct 26, 2026

Commits on Oct 27, 2026