kzoacn/china-advisor
China Advisor Cache This dataset contains cached public web pages, API responses, and extracted text for Chinese university advisor/faculty pages. Current 985 coverage report: 77977 metadata records, 71303 teacher_profile records, and 69186 status-200 teacher_profile records across 39 Project 985 schools. Contents archives/china-advisor-cache-2026-06-20.tar.zst: full cache archive preserving the local data/cache/... layout.… See the full description on the dataset page: https://huggingface.co/datasets/kzoacn/china-advisor.
China Advisor Cache
This dataset contains cached public web pages, API responses, and extracted text for Chinese university advisor/faculty pages.
Current 985 coverage report: 77977 metadata records, 71303 teacher_profile records, and 69186 status-200 teacher_profile records across 39 Project 985 schools.
Contents
archives/china-advisor-cache-2026-06-20.tar.zst: full cache archive preserving the localdata/cache/...layout.archives/china-advisor-cache-2026-06-20.tar.zst.sha256: SHA256 checksum for the archive.metadata/CACHE_FORMAT.md: cache schema and metadata format.metadata/REPORT.md: 985 crawl report.metadata/cache_summary.tsv: per-school cache counts.metadata/cache_summary.json: machine-readable per-school cache counts.metadata/cache_status_summary.tsv: compact status-code notes for notable blocked profile candidates.metadata/errors.tsv: respectful crawl error log.configs/985_sources.yaml: 985 school seed source registry.configs/985_*repair_sources.yaml: targeted central and college-level repair source registries.scripts/crawl_985_cache.py: respectful staged crawler used for the 985 expansion.scripts/repair_985_low_coverage.py: targeted repair crawler for Tsites portals and public API-backed portals.
Restore
From a checkout of this dataset repository:
tar --use-compress-program=zstd -xf archives/china-advisor-cache-2026-06-20.tar.zstThis recreates data/cache/<school>/{pages,text,metadata}/....
Crawl Notes
The cache includes records copied from the original local cache plus respectful 985-wide crawl and repair passes. The crawlers record blocked URLs and do not bypass robots.txt, TLS failures, login, captcha, CC defenses, or other access controls.
