ilia
7b3dfb002a
Some checks failed
CI / Linting (Biome) (push) Failing after 36s
CI / Tests (push) Successful in 5m54s
CI / Type Check (adzuna-extractor) (push) Successful in 1m6s
CI / Type Check (gradcracker-extractor) (push) Successful in 1m9s
CI / Type Check (hiringcafe-extractor) (push) Successful in 1m5s
CI / Type Check (orchestrator) (push) Successful in 1m21s
CI / Type Check (startupjobs-extractor) (push) Successful in 1m4s
CI / Type Check (ukvisajobs-extractor) (push) Successful in 1m4s
CI / Documentation (push) Successful in 1m52s
Adds extractor packages: arbeitnow, ashby, careerjet, fourdayweek,
greenhouse, himalayas, jobicy, jooble, lever, reed, remoteok, remotive,
themuse, usajobs, weworkremotely, workday — each with manifest, package
metadata and README.
Pipeline / shared:
- shared/job-fingerprint: stable hash for cross-source dedup, with tests
- discover-jobs: dedup via fingerprint and richer per-source merging
- jobs repository: fingerprint-aware upsert / lookup
- settings-registry, settings types/routes, demo-defaults: knobs for the
new sources
- shared extractors index: register the new manifests
- location-support, profiles route: small fixes for the new sources
Tooling:
- scripts/smoke-extractors.ts to sanity-check each source locally
- scripts/jobber-cron-{cherepaha,dobkin}.env.example: per-host cron
templates (CHANGEME placeholders only)
- .env.example: documented env vars for the new extractors
- .gitignore: ignore extractors/*/storage/ runtime caches (was ukvisajobs only)
Co-authored-by: Cursor <cursoragent@cursor.com>
workday-extractor
Public Workday career sites via the JSON CXS endpoint
POST {tenantUrl}/wday/cxs/{tenant}/{site}/jobs.
Configuration
Set workdayTenants (or WORKDAY_TENANTS env). Each entry is either:
- A career-site URL we'll auto-parse, e.g.
https://nvidia.wd5.myworkdayjobs.com/en-US/NVIDIAExternalCareerSite - A JSON object with explicit fields:
{
"company": "NVIDIA",
"tenantUrl": "https://nvidia.wd5.myworkdayjobs.com",
"tenant": "nvidia",
"site": "NVIDIAExternalCareerSite",
"locale": "en-US"
}
Multiple entries are separated by newlines or commas. Pipeline searchTerms
are passed as the request searchText.