evolve trial

Inspect one agent attempt. Get its ID from the job’s trial list:

JOB_ID="<job-id>"

evolve job trials "$JOB_ID"

Copy a full trial ID into TRIAL_ID:

TRIAL_ID="<trial-id>"

evolve trial show "$TRIAL_ID"

The detail view includes status, reward, spend, execution details, and an analysis summary when available. Add --json to receive the trial object.

Read the trace

evolve trial trace "$TRIAL_ID" \
  --grep 'permission denied' \
  --tail 20

Option	Meaning
`--type <event-type>`	Select an exact event type, from `1` to `100` characters.
`--grep <pattern>`	Case-insensitive POSIX regular expression over event type and content. From `1` to `512` characters.
`--tail <n>`	Keep the last N matching events. Integer from `1` to `10000`.
`--cursor <seq>`	Resume after this event sequence number.
`-l`, `--limit <n>`	Events per API page. Default `200`, maximum `1000`.

The command reads all available pages and then exits. --limit sets page size, not a total event cap. Use sandbox logs to follow live output. Events are returned in sequence order. With --json, each event is a separate JSON line. To resume later, pass the last event’s seq as --cursor. Invalid regular expressions are rejected. Filters search the stored agent trace. Per-call metered usage added after the trial settles appears only in the unfiltered trace.

Download the trial

evolve trial download "$TRIAL_ID" -o trials/

trials/
└── <task>__<7 characters>/
    ├── config.json
    ├── result.json
    ├── evolve.json
    ├── agent/
    └── verifier/

agent/ stores the trace and agent files. verifier/ stores verifier output. Files are included when their artifacts exist. The command uses the folder name from the archive and prints the saved path. With --json, it returns path and the file count, files.

Option	Meaning
`-o`, `--output-dir <dir>`	Parent folder. Default `trials/`.
`--overwrite`	Replace an existing trial folder. Without it, an existing folder is refused.
`--stream <artifact>`	Print one artifact instead of saving the trial tree.
`--cursor <seq>`	With `--stream trace-parsed`, resume after this sequence.
`--limit <n>`	With `--stream trace-parsed`, set API page size. Default `200`, maximum `1000`.

--stream cannot be combined with --output-dir or --overwrite. --cursor and --limit require --stream trace-parsed; that mode reads every available page, just like trial trace.

Artifact names

Artifact	Content
`trace-parsed`	Parsed trace events.
`trace-stdout`	Raw agent stdout.
`trace-stderr`	Raw agent stderr.
`trace-atif`	Normalized ATIF trajectory.
`verifier`	Verifier output.
`agent-home`	UTF-8 text view of stored agent-home files. Binary files are listed in its manifest.
`filesystem`	Sandbox filesystem archive, as `.tar.gz` bytes.
`trajectory`	Reserved name; native trajectory download is not available through this selector.

evolve trial download "$TRIAL_ID" --stream trace-stdout
evolve trial download "$TRIAL_ID" \
  --stream filesystem > trial-files.tar.gz

With --json, parsed events remain one JSON record per line. Other selectors return:

Artifact	JSON output
Logs, verifier output, or ATIF	`{ "log": "..." }`; missing output returns `log: null`. Withheld output also includes `reason`.
`agent-home`	`{ "files": { "path": "content" } }`, or `files: null` when absent.
`filesystem`	`stream`, byte count `bytes`, `encoding: "base64"`, and base64 `data`.

The reserved trajectory selector returns an API error.

Files, logs, and processes

Use evolve trial files, evolve trial logs, and evolve trial procs to inspect the sandbox. Their commands and flags are in the filesystem reference.

Browse files

List folders, read a file, search text, or download a subtree.

Follow logs

Read agent, verifier, or system output and inspect live processes.

Retry or regrade

evolve trial retry "$TRIAL_ID"
evolve trial regrade "$TRIAL_ID"

Command	Result
`retry`	A new job running the settled trial again.
`regrade`	A new job running only the verifier over eligible stored inputs.

Both commands return a new job ID, or a job object with --json. Use that ID with evolve job show or evolve job trials; the original trial is unchanged. Retry requires a settled trial, even if other trials in its source job are still running. Regrade requires a supported single-step, separate verifier with stored inputs; judge-backed verifiers are excluded. It uses the recorded verifier, not later edits to the task. Uploaded trials cannot be retried or regraded. See Trials.

Stop a trial

Stop selected trials while keeping their job. You can stop only trials from jobs you created. Pass up to 100 IDs; set OTHER_TRIAL_ID to another ID from job trials:

evolve trial stop "$TRIAL_ID" "$OTHER_TRIAL_ID"

The response distinguishes stopped, already-terminal, and missing IDs. To stop an entire dataset’s trials, use job stop. Global options and ID prefixes apply.

Start work

Inspect results

Manage resources