Page through listings

The public listings (/evm/jobs, /evm/providers, /v1/models, /evm/models, /evm/asks, /evm/asks/floors, /evm/allowlist) are paged with limit and offset. A page can end early when the response would pass 4 MiB, so the length of the page alone does not tell you whether you are done.

The loopCopy Anchor Link

Keep paging while either the page is full or the x-vorq-page-truncated header is true:

offset=0
while :; do
  curl -s -D headers.txt "$VORQ/evm/jobs?state=Open&limit=1000&offset=$offset" > page.json
  jq -c '.jobs[]' page.json            # consume the page
  n=$(jq '.jobs | length' page.json)
  truncated=$(grep -i '^x-vorq-page-truncated:' headers.txt | tr -d '\r' | awk '{print $2}')
  next=$(grep -i '^x-vorq-next-offset:' headers.txt | tr -d '\r' | awk '{print $2}')
  [ "$n" -eq 1000 ] || [ "$truncated" = "true" ] || break
  offset=${next:-$((offset + n))}
done

• limit is 1–1000 (default 100); offset is 0–1 000 000.
• x-vorq-next-offset is present only on a truncated page. Otherwise the next offset is offset + returned.

Narrow instead of paging deepCopy Anchor Link

Offsets past one million are refused. For a large job book, filter rather than page: /evm/jobs takes state, model, provider, owner, min_rate_in, min_rate_out and posted_before, and order=newest puts the latest jobs first. See GET /evm/jobs.

Rows move between pagesCopy Anchor Link

Each page carries as_of_block, the index block it was read at. The index keeps moving between pages, so rows can shift. On the job book, pass the first page's as_of_block as posted_before so that jobs posted while you page do not push rows onto the next page. Jobs can still change state between pages.

GET /v1/batches pages differently, by limit and after; see GET /v1/batches.