All notable changes to this Actor are documented here.
- Closed jobs are no longer re-reported (and re-charged) on every run. Once a closed posting has been pushed, its key is dropped from the monitor snapshot; previously the same closed job was emitted again on every subsequent run for up to 90 days. If it reappears on the board it is correctly "new" again.
- New-job mode no longer loses postings: a new job cut off by
maxItems or the charge budget now stays "new" for the next run. Previously it was recorded as seen without ever reaching the dataset, so onlyNewJobs runs would never deliver it. (Jobs excluded by the user's own title/location filters are still recorded — deliberate, unchanged.)
- Greenhouse embed board URLs (
boards.greenhouse.io/embed/job_board?for=<token>) now resolve to the right company token instead of the literal token embed.
- Percent-encoded tokens pasted as URLs (e.g.
jobs.ashbyhq.com/My%20Company) are now decoded before the API URL is built, instead of being double-encoded into a 404.
- HTML-entity decoding in plain-text descriptions now decodes
& last, so double-escaped text (e.g. &lt;) is decoded exactly once and can no longer reintroduce markup into descriptionText.
- When the pay-per-event budget limit is reached, the actor now actually stops emitting items (boards are still fetched so new/closed-postings history stays accurate).
- "Last seen" history entries with a missing or corrupt timestamp are now pruned instead of being retained forever.
- New closed-postings detection:
includeClosedJobs input emits a row (isClosed: true) for jobs seen in a previous run but no longer listed on a board actually checked this run. The persisted per-account history now stores a small snapshot (title/department/location/jobUrl) per job, not just a timestamp, so closed rows carry useful context instead of just an ID.
- Guarded against false positives: a board dropped from input, or one that fails transiently, never causes its previously-seen jobs to be reported as closed — only boards successfully checked in the current run are considered (verified with a live test: dropping a board from input produced 0 false "closed" rows for it).
- New "new postings" mode: every row gets an
isNew field, and the onlyNewJobs input filters output to just jobs not seen in a previous run. History is tracked per Apify account in a named key-value store, independent of the in-run maxItems/title/location filters, with a 90-day rolling retention.
- Renamed the actor to lead with the ATS platform names for discoverability, since "ATS Jobs Scraper — Greenhouse, Lever & Ashby" is already used verbatim by two other Store listings.
- Added Ashby as a third supported ATS platform (
ashby:token or https://jobs.ashbyhq.com/…), alongside Greenhouse and Lever, using Ashby's public job board API.
- New
compensationSummary field, populated from Ashby's published salary/equity ranges (null for Greenhouse/Lever).
Initial release.
- Full extraction with a flat, CSV/Excel-friendly dataset schema and a Store output table view.
maxItems cap enforced globally (safe under concurrency), plus input validation with clear, fail-fast errors.- Pay-per-event pricing:
actor-start + per-result-item charging (charged only after an item is produced).
- Apify Proxy support, automatic retries with backoff, and graceful per-item degradation (one bad page never crashes the run).
- Resumable & migration-safe via
Actor.useState (no double-charge on platform migration).
- Input/redirect SSRF guards (block localhost, private/reserved IPs, cloud-metadata,
file://) with DNS-rebinding resolution checks where user URLs are fetched.
- ReDoS-safe extraction patterns and response-body size caps.
- Push-failure rollback so a failed push never silently loses items or mischarges.
