curl and jq: count Sume jobs by status in one command

Pipe GET /v1/jobs?limit=100&scope=workspace into jq and group the data.jobs array on status. The list returns up to 100 jobs a page, each with one of five statuses, so a single command shows how many are queued, processing, completed, failed or canceled. The third line of the script prints a warning when the response carries next_cursor, which means more pages remain.

What the list call accepts

The parameters below come from Sume's OpenAPI spec. The response is data.jobs[] plus data.next_cursor, which is absent on the last page and which you pass back as starting_after. Do not build a cursor from the last job's id.

The command

Set SUME_API_KEY and run it. It prints one tab-separated line per status.

To count a single status without grouping, add &status=queued to the URL and read .data.jobs | length.

curl -s "https://api.sume.com/v1/jobs?limit=100&scope=workspace" \
  -H "Authorization: Bearer $SUME_API_KEY" \
| jq -r '(.data.jobs | group_by(.status) | map("\(.[0].status)\t\(length)") | .[]),
          (if .data.next_cursor then "more pages: starting_after=\(.data.next_cursor)" else empty end)'

What the counts do not mean

An API key reads only the jobs its own member created, and other jobs return 404, so the counts describe your key's member, not the whole workspace, even with scope=workspace. The counts cover one page. Over 100 jobs, loop on next_cursor or filter by status.

The numbers are a snapshot. A job can move from queued to processing between two calls. Use them to eyeball a backlog, and use generation_limits in a submit response when you need admission headroom.