Append items to an open job

The /v1/scrape request object as accepted by batch, minus url. Field names, types and validation are exactly ScrapeRequest's (enforced by a parity test); each property below references it. Every item is merged as job-level value, overridden by the item's value (an explicit false on an item wins), with the project's dashboard defaults underneath both. Validation happens for every item at submission.

What the batch worker executes: each item is one fetch or one render, and the document is returned as response_format (html by default, json, markdown or text). A field is either honoured on a batch item or refused at submission, never accepted and skipped. Refused with 400 INVALID_PARAMETER, naming every offending field, with nothing queued or charged: actions, screenshot and its options, network_capture, extract, extract_preset, ai_extract, autoparse, links, session_id, cache/cache_ttl, original_status, allowed_status_codes, parse_pdf=true, mode=auto, response_format=pdf, and a non-GET method with js_render or stealth. A value that asks for nothing (screenshot: false, an empty list) is accepted. A refusal caused by a project default says so in detail. Use /v1/scrape for those fields.

A batch item is priced when it is accepted, so a bot challenge never escalates it to stronger rendering or retries it on a different exit the way it can a /v1/scrape request with your own proxy.