Skip to main content
Set an API deadline for the work, and a longer client timeout for the response. If your client gives up first, the call may still be running.

Deadlines

timeout is in milliseconds unless the table says seconds. A call that times out costs nothing. A short scrape timeout that runs out answers 504 sourceTimeout with retryAfterSeconds: 5. Retry with a longer timeout.

Background jobs

Crawl, batches, extract and agent answer at once with an id. Read the result with their status calls, so your client never waits for the work. A browser task works the same way: read it with POST /v1/browser/task/status. Results are kept for 24 hours: a scrape you read again with GET /v1/scrape/{id}, a crawl, a batch and an extract. After that the id answers notFound. Zero data retention shortens this.

Your client timeout

Most calls answer within 15 seconds. A slow page, a long PDF or a transcript can take a minute, and a scrape with the default timeout can take 60 seconds. Set your client timeout to at least 65 seconds, or longer than the timeout you send plus a few seconds. The SDKs wait 70 seconds for each attempt. If a write may have run and cannot be replayed safely, the SDK stops and marks the outcome unknown. Before you resend after a timeout, check whether the first call finished. If the error carries an idempotency key, resend with the same key and body to recover the original result. Retry rules are in Troubleshooting.