Runs

A run is one execution of an actor with an input. It has a status, a live log, its own storages and a usage record. Find your runs under Actors → Runs in the Console, or with GET /api/v3/runs.

Statuses

Status	Meaning
READY	Accepted and waiting for a machine to pick it up.
RUNNING	The actor's container is running.
SUCCEEDED	The actor finished without error.
FAILED	The actor exited with an error. The log usually says why.
TIMED-OUT	The run reached its timeout and was stopped.
ABORTED	Someone stopped it before it finished.

SUCCEEDED, FAILED, TIMED-OUT and ABORTED are final. Results the actor pushed before the end stay in its dataset whatever the status.

Each run also records where it came from in origin: WEB (the Console), API, SCHEDULER, WEBHOOK or STANDBY. A run started by a schedule shows Schedule: and the schedule's name, and the API returns it in schedule.

Lifecycle

1. Start. POST /api/v3/actors/:actor/runs or POST /api/v3/tasks/:task/runs, the Start button, a schedule, an integration or an AI app.

2. Checks. Before a run is accepted, Lantfly checks:

   Check	Refused with
   The input matches the actor's input schema	422 validation_failed
   The account owner's email is verified	403 email_not_verified
   The account has usage left (plan usage, credit or overage room)	402 not_enough_usage_to_run_paid_actor
   The account is not on billing hold	402 billing_on_hold
   Runs waiting or running are within your plan's concurrent run limit	429 concurrent_runs_limit_exceeded

3. READY. The run gets a default dataset, key-value store and request queue, and a run token.

4. RUNNING. The actor reads its input, pushes results and writes records, and its log streams in.

5. Finish. The run ends in a final status. Lantfly records usage, fires webhooks, checks monitors and sends notifications.

Wait for a run

wait_for_finish (0 to 60 seconds) on POST /api/v3/actors/:actor/runs and POST /api/v3/tasks/:task/runs keeps the request open until the run finishes or the time is up, then returns the run. For longer runs, poll GET /api/v3/runs/:run, or get a webhook when it ends.

The run page

Tab	What it shows
Output	The dataset items, as a table or as the actor's own dataset views, with Export. Links from the actor's output schema are listed too.
Log	The live log. Download the full log saves all of it.
Live view	A web page the actor serves while it runs (see Live view).
Input	The input the run started with.
Key-value store	The run's records, such as INPUT and OUTPUT.
Storage	The IDs of the default dataset, key-value store and request queue.
Usage	Compute units, results and charged events, with their cost.

Logs

GET /api/v3/runs/:run/log?from=0 returns { chunk, offset, finished }. Each call returns at most 1 MB; call again with from set to the offset you received until the run has finished and chunk comes back empty. Secrets (actor environment variables marked secret) are masked. A run keeps at most 50 MB of log; past that the log is truncated.

Read a run log

curl -s "https://lantfly.com/api/v3/runs/$RUN_ID/log?from=0" \
  -H "Authorization: Bearer $LANTFLY_TOKEN"

Abort

Abort (or POST /api/v3/runs/:run/abort) stops a READY or RUNNING run. A READY run is aborted at once; a RUNNING one stops within moments. Results pushed so far stay in the dataset.

Run again and resurrect

Action	What it does
Run again	Starts a new run with the same input and options, with new storages. It is disabled when the old input cannot be read. A run that came from a task runs again through the task.
Resurrect	Starts the same run again on its own storages, as long as they have not been removed by retention. Results it already pushed stay and new ones are added. You may change memory, timeout and build.

Resurrect goes through the same checks as a new run (usage, concurrent runs), is billed for the extra time only, and fires the ACTOR.RUN.RESURRECTED webhook event. The run's runtime adds up across resurrections.

Resurrect a run with more memory

curl -s -X POST "https://lantfly.com/api/v3/runs/$RUN_ID/resurrect" \
  -H "Authorization: Bearer $LANTFLY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{ "options": { "memory_mb": 2048 } }'

Resurrect answers 422 when the run has not finished.

Reboot

Reboot (or POST /api/v3/runs/:run/reboot) restarts the container of a RUNNING run. The actor starts over on the same storages; anything it kept only in memory is lost. It answers 422 when the run is not running.

Metamorph

An actor can turn its own run into a run of another actor with POST /api/v3/runs/:run/metamorph (target_actor_id, input, optional build). The run keeps its storages; the new input is stored under a key INPUT-METAMORPH-1, INPUT-METAMORPH-2 and so on. The target must be a public actor or one of the owner's. The run keeps the prices it started with, so metamorphing into another account's paid actor is refused (422). Past metamorphs are listed in the run's metamorphs.

Status message

While it runs, an actor can show a short status (for example "Page 12 of 40") with PATCH /api/v3/runs/:run { status_message, is_status_message_terminal }. A message marked terminal stays after the run ends.

Live view

Some actors serve a web page while they run, such as a screenshot stream or a small control panel. When a run is RUNNING and has one, the run page shows a Live view tab, and GET /api/v3/runs/:run returns a container_url you can open in a browser (it carries a short-lived token). Use Open in new tab if the embedded view stays blank. container_url is null before the run starts, after it ends, and for actors without a web server.

Find runs

GET /api/v3/runs lists runs newest first and takes these filters:

Parameter	Meaning
actor_id	Runs of one actor
task_id	Runs of one task
schedule_id	Runs a schedule started
status	One or more statuses, comma-separated
from, to	Created at or after / at or before (a date or date-time; a date in to means that whole day)
page, per_page	Paging, up to 100 per page

The Console Runs page has the same filters, and they stay in the URL so you can share a filtered view.

Bulk actions

POST /api/v3/runs/bulk with { "ids": [...], "action": "abort" } or "resurrect" acts on up to 200 runs at once. Runs that cannot take the action (already finished, still running, storages gone) are listed in failed; the others go on. See Pagination and Rate limits for limits on list and bulk calls.

Limits

Limit	Value
Concurrent runs (waiting or running)	Set by your plan; see Billing
Memory per run	128 MB to 8,192 MB by default
Timeout per run	10 seconds to 24 hours by default
wait_for_finish	Up to 60 seconds
Ad-hoc webhooks per run	10
Log page	1 MB per call; 50 MB per run
Bulk actions	200 runs per request
Run storage kept	Set by your plan; name a storage to keep it (see Storage)