Oxylabs has two Web Scraper API endpoints and they are not interchangeable: realtime.oxylabs.io/v1/queries answers synchronously, data.oxylabs.io/v1/queries is the async Push-Pull path with callbacks and S3/GCS delivery. Long or heavy jobs on the realtime endpoint time out; polling-style code on the data endpoint waits for an inline answer that never comes. Pick realtime for quick inline results, data plus callback_url for everything else - and remember Push-Pull results expire after at least 24 hours, so fetch them promptly.

Context: Oxylabs' official agent skill (oxylabs/agent-skills, web-scraper-api): two different endpoints with different semantics. POST https://realtime.oxylabs.io/v1/queries returns an immediate response - for synchronous jobs. POST https://data.oxylabs.io/v1/queries creates Push-Pull jobs with callbacks and cloud storage - asynchronous. Use realtime when you need the answer now in the response body; use the data endpoint when the job is long, you want a callback_url ping on completion, or results go straight to S3/GCS. Mixing them up means either sync timeouts on long jobs or polling an endpoint that never returns inline.

## Matched source
Source: Source: https://github.com/oxylabs/agent-skills/blob/HEAD/skills/web-scraper-api/SKILL.md
Original query: "Oxylabs realtime vs Push-Pull: sync endpoint for quick answers, async endpoint for long jobs and callbacks"
Key terms: answers, async, callbacks, endpoint, jobs, long, oxylabs, pull, push, quick, realtime, sync