Grok AutomationAdd to Chrome
08 · Operations · 5 min read

Queue, retries, delays, and downloads

Once you've shipped a few batches, these are the controls that decide whether a 200-prompt run lands smoothly or chews through your quota.

How the queue runs

Grok Automation’s runner is intentionally serial — one prompt at a time. The Concurrent field in Settings → Queue is fixed at 1 and disabled in the UI.

Why? Grok’s rate limiter punishes parallelism much harder than it punishes pacing. A 10-prompt batch with a 2-second gap between submissions almost always finishes; the same 10 prompts fired in parallel often gets one through and rate-limits the rest. The runner trades wall-clock time for completion rate.

What you can tune lives on the Settings page (gear icon, top right of the side panel), under the Queue card.

The three queue knobs

Screenshot pending Settings page Queue card with four fields: Concurrent (disabled), Retries per prompt, Delay min, Delay max
Concurrent is locked. The three you actually tune are Retries per prompt, Delay min, and Delay max.

Retries per prompt

How many times the runner re-submits a single failed prompt before giving up and moving on. Default: 2.

Set to…When
0You’d rather see failures fast and re-run them yourself. Useful when iterating on prompt wording — you don’t want retries hiding a bad prompt.
2 (default)Normal use. Handles transient grok.com hiccups without burning quota on permanently-broken prompts.
4–5Long unattended overnight batches where you’d rather wait than re-queue tomorrow.

Retries cost real Grok quota — each retry is a full new generation. The default 2 is a deliberate balance.

Delay min (s) / Delay max (s)

The runner waits a random amount of time between prompts, picked uniformly between Delay min and Delay max. Defaults: 2 and 5 seconds.

Randomness here matters: a fixed 3-second gap looks like a bot to Grok’s rate detector. A 2–5s random gap looks like a busy human. Keep the spread.

ScenarioDelay minDelay max
You’re babysitting, want it fast13
Default25
Long overnight batch (200+ prompts)512
You got rate-limited last run820

The “starting in X” countdown that appears on the next-up prompt row is this delay in action.

When something interrupts the run

Three things can stop a queue mid-flight:

  1. You clicked Cancel on the Current run card.
  2. Grok rate-limited the account (the rate-limit modal pops up).
  3. You ran out of free quota (the quota-exhausted modal pops up).

In all three cases, the runner saves a snapshot of the queue — which prompts succeeded, which failed, which never ran. That snapshot lives in chrome.storage and survives browser restarts.

The next time you open the side panel in the same mode, the big orange button shows:

Continue · 12 unfinished →

Click it and only the un-run prompts execute, with all your original settings (mode, aspect, references) preserved.

Continue is per-mode. If you ran a Text-to-Image batch and now you’re in Text-to-Video, the Continue button only shows up when you switch back to Text-to-Image. Each mode has its own snapshot slot.

If you want to abandon the snapshot instead, the Current run card has a Discard button. The next run starts fresh.

Downloads, in detail

The Downloads section on the main panel controls where finished files land. Toggling Auto-download off keeps generations in the Grok tab without saving — useful if you want to manually pick which ones to keep.

Folder

Subdirectory under your Chrome Downloads location. Defaults to grok-auto. Use one folder per project (q4-thumbnails, book-cover-v3) to keep work separated.

Filename prefix

Prepended to every filename. Pairs well with project folders for double-tagging: folder=q4-thumbnails + prefix=v2- produces q4-thumbnails/v2-1_neon-alley.png.

Use the prompt text as the filename

When on, filenames derive from the prompt itself (lightly sanitized). When off, files are sequential — prefix-001.png, prefix-002.png, etc.

SettingBest for
OnSearchable archives. Files self-document.
OffSequenced batches you’ll edit by index. Less typing in a video editor.

History — see what you’ve run

The History card on Settings shows how many run snapshots are currently stored locally. The number includes both completed and interrupted runs.

Click Clear completed history to wipe finished snapshots only — interrupted runs (the ones that still have a Continue offer) are preserved.

A common pitfall: prompts changed mid-run

If you edit the prompt textarea while a run is in progress, the textarea is read-only during the run — but if you cancel and then edit, the snapshot detaches and Continue disappears. Next click of Run will start a fresh batch from your edited prompts. This is intentional: editing means you’ve changed your mind, so the old queue is no longer what you wanted.


Grok Automation is an independent browser extension for Grok users. Not affiliated with xAI.