jev batch

Run a command over many rows with a concurrency pool. One JSON record per row, in input order.

jev batch <command> --input <ref> [--output <path>] [--concurrency <n>] [--fail-fast] -- <flags for that command>

Batchable: classify, screen, extract, ask, verify, find, rerank, route.

When to use

Anything over a handful of items: a backlog to label, a folder of emails to extract from, a list of claims against one spec, a set of queries against one candidate list.

Options

Option	Meaning	Default
<command>	The command to run per row	required
-i, --input <ref>	@file or -. Plain lines (one text per line), or JSONL objects with text (or state) and an optional id. Extra fields pass through as meta.	required
-o, --output <path>	Write records to a file instead of stdout	stdout
--concurrency <n>	Parallel requests	4
--fail-fast	Stop scheduling new rows after the first error	off
-- <flags>	Everything after -- goes to the sub-command, minus its text argument, which each row supplies

For verify the row is a claim and --evidence is shared; for find and rerank the row is a query and the candidates are shared.

Output

One line per row: {"index", "id", "ok", "failed", "result" | "error", "meta"}. failed means the sub-command's --fail-on matched for that row. A summary goes to stderr.

Exit 1 if any row errored, 2 if any row matched --fail-on, else 0.

Example

jev batch classify -i @issues.jsonl -o labeled.jsonl -- -l bug,feature,question,docs --other --fail-on review
jq -r 'select(.failed) | .id' labeled.jsonl      # rows a human should look at

jev batch screen -i @pages.jsonl --concurrency 8 -- --purpose "extract pricing"
jev batch verify -i @claims.txt -- --evidence @spec.md --fail-on contradicted,unsupported

# Build JSONL from files
for f in inbox/*.eml; do jq -n --arg id "$f" --arg text "$(cat "$f")" '{id:$id,text:$text}'; done \
  | jev batch extract -i - -- --want sender=email:the sender --want reply_by=date:the reply deadline