Managing & debugging automations
Open the Automations drawer to manage every automation, then open a card and read its Activity log under the canvas to debug almost any "why didn't this fire?" question.
The card list
Each automation card in the drawer shows a bolt icon tinted with the automation's color, the name, and a one-line trigger summary like Gmail · Subject: "invitation to bid" → Bid mailbox or Drive · Vendor Onboarding → Vendor mailbox. It also shows the relative time of the last run (23 minutes ago), or, on failure, the trimmed error message, plus a toggle to enable or disable. When something blocks it, an amber Auto-paused chip and an explanatory banner appear. Clicking the card body opens the flow builder.
Edits and toggles are restricted to the user who created the automation. Workspace owners can additionally delete automations whose powering user has been removed.
Inside the editor
Open a card to see the three flow nodes (Trigger, Destination, Extract Data), each click-to-expand, an Activity log under the canvas, and a footer with Last triggered, Status, Powered by, plus Backfill and Delete.
Backfill replays a Gmail automation over a historical date range, useful right after creating it to catch messages that predate the rule. It processes 50 messages at a time (the button becomes Continue · 50 more with a progress bar mid-range), and an On duplicates choice controls already-imported files for that run: Skip (default), Overwrite (re-extract in place, handy after a template fix), or Rename (numbered copy). Delete is permanent: disable first if you might revisit.
The source link is annotation, not file routing: files land at the case's storage path; the linked document tells the workspace which automations feed which documents. With Extract Data on and a template picked, the automation links to that template's document, and every ingested file is tagged to match it (the join key the picker ranking, send-template attachment, and case-detail visibility all read). One linked document is used automatically; with several, pick the Incoming document type. Turn extraction off and you pick the Document directly on Destination instead. Mailbox always lives on Destination.
Debugging: triage in order
- Did the trigger match? No Activity entry means it never fired. Check filters (Gmail label match is exact; all set filters must match; Drive matches the folder you specified, not subfolders), and check the Gmail / Drive toggle rows in Settings → Account for the underlying connection.
- Skipped on purpose? A Skipped entry says why, usually the file was already imported from the same source (duplicates skip automatically).
- Category or template resolve? An Error entry names the broken step. Ingestion fails closed when neither an Extract template nor a Destination category resolves: uncategorized files would break downstream filters.
- Extraction fail? The detail line names the cause: a malformed or unsupported file, an AI provider timeout, or a template that doesn't fit the document.
- Paused? A toggled-off card or Auto-paused chip. The banner names the exact recovery. A paused automation writes no Activity entries.
Activity entries are timestamped and color-coded Processed (file count plus whether extraction ran), Skipped, or Error (summary names the step). Recent runs only: older history rotates out, so reproduce old issues with a fresh test message or file.
Pause reasons & fixes
Fix the underlying cause before re-enabling: flipping the toggle without reconnecting just re-pauses on the next run.
| Banner | What it means |
|---|---|
Paused manually | Someone turned the toggle off. |
OAuth expired, reconnect Gmail | Reconnect via the Gmail toggle row in Settings → Account. |
OAuth expired, reconnect Google Drive | Same, for Google Drive. |
Connected user removed from workspace | Reassign the powering user, or have an owner delete it. |
Wrong mailbox or document? Mailbox lives on the Destination node. The document is on Destination with extraction off; with extraction on it is the Incoming document type on the Extract node.