README.md

---

pretty_name: Claude Code Benchmark Suite
language:
  - en
  - zh
tags:
  - benchmark
  - cli
  - evaluation
task_categories:
  - text-generation
size_categories:
  - n<1K
source_repository: https://github.com/anthropics/claude-code
restored_repository: https://github.com/oboard/claude-code-rev
upstream_repository: https://github.com/anthropics/claude-code
---


# Benchmark Suite / 基准测试套件

Professional benchmark harness for this restored Claude Code repository.

面向本还原版 Claude Code 仓库的专业级 Benchmark 框架。

Original upstream repository / 原版仓库地址：  
https://github.com/anthropics/claude-code

Restored source repository / 还原版源码仓库地址：  
https://github.com/oboard/claude-code-rev

## Commands / 命令

- `bun run bench:smoke` - quick hybrid pass for local validation  
  快速混合模式验证（本地冒烟）
- `bun run bench:full` - full hybrid suite  
  完整混合模式基准
- `bun run bench:offline` - deterministic offline-only run  
  仅离线可复现模式
- `bun run bench:online-sample` - online-only sampled run (requires credentials)  
  仅在线抽样模式（需要凭据）
- `bun run bench:compare --baseline=<path> --current=<path>` - compare two JSON reports  
  对比两份 JSON 报告

## Scenario Mapping / 场景映射

- `B01` version fast path / 版本快路径
- `B02` startup phase profiling / 启动分段剖析
- `B03` bare-mode delta / bare 模式差分
- `B04` command catalog load / 命令目录加载
- `B05` headless single round (offline fixture + online sample) / 无头单轮（离线夹具 + 在线抽检）
- `B06` headless streaming format and timing checks / 无头流式格式与时延检查
- `B07` slash/queue batching correctness / 斜杠命令与队列批处理正确性
- `B08` tool orchestration pipeline behavior / 工具编排流水线行为
- `B09` restoration health gate via `dev-entry` / 通过 `dev-entry` 的恢复健康门禁

## Outputs / 输出

Reports are written to `benchmark/results/`.

报告会写入 `benchmark/results/` 目录：

- `benchmark-<run>-<timestamp>.json`
- `latest.json`
- `latest-summary.md`
- `comparison.md` (when running compare / 运行 compare 时生成)