用 Claude Code 规划你的高考。 gaokao.ha7ch.com
把下面这段 prompt 粘进 Claude Code / Codex / Cursor：
跑 `npx gaokao-pro@latest help` 把命令摸清楚，然后帮我规划 2026 年的高考志愿。
先问我：分数（估分 / 模考分 / 高考分都行，标清楚是哪种）、省份、选科组合、目标专业方向或职业兴趣、偏好（目标城市 / 是否限定 985/211 / 学费预算）。如果给的是估分或模考分，先跑 rank-tables 确认该省收录的年份科类，参考 2024-2025 历年一分一段做粗估位次并向我说明数据年份；等高考真实分数出来、且 2026 当年一分一段已收录后再精算。
每条推荐都用 CLI 拉真实数据支撑——查历年最低分、跨校搜专业、把分数换算成位次区间。
要快：先用 `recommend` / `top`（离线）筛出候选名单，再用 `batch 校1,校2,...` 一次并行拉多校真实录取数据，不要对每所学校单独反复调 `actual`。同一查询会被缓存，重复查近乎瞬时。
npx gaokao-pro@latest helpOr globally:
npm install -g gaokao-pro(Curl|bash installer is also still available:
curl -fsSL https://raw.githubusercontent.com/HA7CH/gaokao-pro/main/install.sh | bash)
gaokao-pro is a CLI + MCP server that grounds an AI conversation in
official Chinese college-admissions data. Claude drives the flow; the
CLI is the data spine.
Tools Claude can call (via Bash or MCP):
| Verb | What it does |
|---|---|
recommend |
冲 / 稳 / 保 buckets for your score in a province (offline, 2,400+ schools) |
top |
Top-N best schools your score can reach |
find |
Search majors across schools — e.g. all 985 schools recruiting 计算机 |
school |
University metadata: 985 / 211 / 双一流 / 学科评估 / 排名 |
plan |
Forward-looking admission plan (year × province × school) |
actual |
Backward-looking actual admissions: 最高/最低/平均分 + 最低位次 |
batch |
Scan many schools' actual data in one parallel call (fast 冲稳保 triage) |
scores |
Historical min-score time series for a (school, province) pair |
rank |
score ↔ 全省位次 via official 一分一段表 (30 省 2024+2025 双年份) |
provinces |
List 31 provinces with their 新高考 reform mode |
cache |
Inspect/clear the gaokao.cn response cache (24h TTL; repeat calls instant) |
mcp |
Run as MCP server — claude mcp add gaokao-pro -- npx -y gaokao-pro mcp |
Data sources:
static-data.gaokao.cn(中国教育在线 / 掌上高考 static JSON tier) — no auth, no sign, no rate limit. Powersschool/plan/actual/scores/find/ the offline index forrecommend/top.cli/data/yifenyiduan/— extracted 一分一段表 JSON, per province per year per track. 30 省 2024+2025 双年份已收录（西藏官方不发布逐分表）。
gaokao-pro/
├── cli/ # npm package
│ ├── src/
│ │ ├── index.ts # CLI router + help
│ │ ├── gaokao-cn.ts # static-data.gaokao.cn client
│ │ ├── recommend.ts # 冲/稳/保 algorithm (offline)
│ │ ├── top.ts / find.ts # top-N + cross-school search
│ │ ├── rank-table.ts # 一分一段 loader (score ↔ rank)
│ │ ├── mcp.ts # stdio MCP server
│ │ ├── format.ts # TTY table rendering
│ │ ├── index-loader.ts # gunzip + cache the school index
│ │ ├── probe.ts # rebuild the school index
│ │ ├── codes.ts # 31 province codes + 新高考 reform map
│ │ └── provinces/ # province-bureau fallback adapters
│ ├── data/
│ │ ├── school-index.json.gz # 3,145-school corpus (gzipped)
│ │ └── yifenyiduan/ # 一分一段 (30 省 × 2024/2025)
│ └── test/smoke.ts # live API smoke (10 checks)
├── src/ # gaokao.ha7ch.com landing page
└── docs/ # data-sources scan + schema notes
pnpm install
pnpm dev # gaokao.ha7ch.com landing on :3000
pnpm -C cli dev recommend --score 660 --province henan --subjects 物理,化学,生物 --985
pnpm -C cli test # 10 smoke checks against live API + local data
pnpm -C cli probe # rebuild cli/data/school-index.json.gz
pnpm -C cli build # tsc → cli/dist/The infrastructure is in place — adding a new province is a JSON drop:
- Get the 一分一段表 PDF/Excel from the 省考试院 (see
docs/data-sources.mdfor URLs). - Extract rows in this shape:
{ "province": "henan", "province_name": "河南", "year": 2024, "track": "physics", "source": "河南省教育考试院 (heao.com.cn)", "count": 547, "rows": [ { "score": 700, "count": 12, "cumulative": 12 }, { "score": 699, "count": 18, "cumulative": 30 } ] } - Save to
cli/data/yifenyiduan/{province-pinyin}-{year}-{track}.json. rank/rank-tables/ MCP pick it up automatically.
Tracks: combined (3+3 provinces), physics / history (3+1+2), science / liberal (老高考).
MIT. Part of HA7CH — sibling of job-pro and cv-pro.