Overview

This page covers the cloud-storage nodes:

All of them are OAuth-backed and rely on server-side execution for real file access.

Configuration / Settings

Node	Setting	Type	Required	Details
`trigger`	Google account / `connectionId`	string	Yes	Required to watch a Drive folder.
`trigger`	Watched folder / `folderId`	string	Yes	Only direct children are reliably watched in the current implementation.
`trigger`	`allowedExtensions`	comma-separated string	No	Filters new files by extension.
`trigger`	`recursive`	boolean	No	Present in the UI, but current watcher logic is still effectively direct-child only for folder-scoped watches.
`googleDrive`	`mode`	`download` or `upload`	No	Defaults to `download`.
`googleDrive`	`fileId`	string	Required for download	Picks the Drive file to fetch.
`googleDrive`	`folderId`	string	Optional for upload	Upload target. Empty means Drive root.
`dropbox`	connection settings	string	N/A	The node exists, but executor support is still incomplete.

trigger receives Drive change events from the watcher service and downloads the changed file before passing it downstream.
googleDrive in download mode does not require upstream input.
googleDrive in upload mode expects upstream Blob or File objects.
dropbox currently returns a failed status in server execution and is not production-ready.

Node	Output shape	Notes
`trigger`	`File[]`	The watched file is downloaded and returned as an in-memory file object.
`googleDrive` download	`File[]`	One downloaded file wrapped in an array.
`googleDrive` upload	`Blob[]`	The original input files are passed through after upload.
`dropbox`	passthrough failure	The executor reports that Dropbox workflow execution is not fully implemented.

Important metadata:

Drive watcher inputs include fileId, fileName, mimeType, size, webViewLink, downloadUrl, and connectionId before download.
Upload mode records uploads metadata with the Drive API result.

The initial watcher run establishes a page token and does not trigger the workflow.
Extension filtering is string-based. Use values like .pdf,.docx.
Recursive watch is currently documented as a limitation because the watcher still checks direct parents for folder-scoped runs.
Client-side execution of these nodes is intentionally blocked because OAuth tokens are handled on the server.