Reference
Schedules, statuses, supported file types, limits and processing cost for integration flows.
The details behind integration flows, collected for looking up rather than reading through.
Schedules
| Frequency | When it runs |
|---|---|
| Manual only | Never on its own; only when someone chooses Run now |
| Every day | Once every 24 hours |
| Every 3 days | Once every 72 hours |
| Every week | Once every 7 days |
A schedule only applies to an active flow. Drafts and paused flows never run automatically, whatever frequency they carry. One run happens at a time per flow, so a scheduled run is skipped if the previous one is still going.
Statuses
A flow is a draft when first created and runs nothing; active once it is on its schedule; paused when someone stops it; and in error when it can no longer read its folder, usually because the connected account lost access or was disconnected. A flow in error carries the reason and does not run again until fixed and reactivated.
A run is running, completed, or failed. A failed run means the run itself could not proceed, such as being unable to list the folder, rather than individual documents failing.
A document moves from discovered, to downloading, to parsing, to parsed, to extracting, to extracted. Failed means that document could not be processed; the run carries on with the others. Failed documents are retried on up to three later runs before being left alone, and a document edited in the folder gets a fresh set of attempts.
An extraction is what you act on in the Review tab:
| Status | Meaning |
|---|---|
| Needs field mapping | Extracted, but the flow has no entity mapping so it cannot be approved yet |
| Pending review | Ready for someone to check and approve |
| Approved | Approved and being written |
| Synced | Written onto an entity record |
| Sync failed | Approval could not be written; fix the cause and approve again |
| Rejected | Discarded by a reviewer and never written |
Supported documents
| Type | Extensions |
|---|---|
.pdf | |
| Word | .docx |
| Excel | .xlsx, .xlsm |
| Text and data | .txt, .csv, .tsv, .md, .json |
Google Docs, Sheets, Slides and Drawings are exported to PDF automatically before being read, so they can sit in a watched folder as they are. Other Google native types, such as Forms, cannot be exported and are skipped.
Anything else in the folder is marked as failed with the reason, and the rest of the run continues. Legacy .doc and .xls files are not supported; resave them in the current format.
Scanned documents are not read
Reading a document means reading its text. A scan or photograph has no text to read, only an image of it, so it fails with a note that it needs OCR. This applies to PDFs produced by a scanner, which look identical to ordinary PDFs until you try to select text in them.
Limits
| Limit | Value |
|---|---|
| Documents per scheduled or manual run | 200 |
| Documents per sample run | The flow's sample size, 1 to 50 |
| Maximum file size | 50 MB |
| Retries for a failed document | 3 |
A folder holding more than 200 documents is not a problem; a run takes the first 200 and later runs continue through the rest.
Who can do what
| Action | Owner | Admin | Member | Guest |
|---|---|---|---|---|
| Enable a provider for the organization | Yes | Yes | No | No |
| Connect or disconnect their own account | Yes | Yes | Yes | No |
| Create, edit, delete or run a flow | Yes | Yes | No | No |
| View flows and their runs | Yes | Yes | Yes | No |
| Review, edit and approve extractions | Yes | Yes | Yes | No |
Reviewing is deliberately open to members, since checking what a document said is the part most likely to need someone with the relevant knowledge rather than the relevant role.
A flow is shared rather than personal, so everyone who can see flows sees the same list. What a flow can read is bounded by the folder access of the account it was connected with, whatever the reviewer's own role. See Roles and permissions.
Processing cost
Reading a document costs 2 credits per page, charged at the point it is read and before anything is extracted. Approving, rejecting, editing and re-approving are all free, so reviewing carefully costs nothing extra.
Page counts follow the document: a PDF is charged by its real page count, a spreadsheet by its number of sheets, and Word or text files as a single page. Charging is per read rather than per document, so a document read again in a later run is charged again. This is worth knowing while tuning a flow, since repeated sample runs over the same folder each cost credits. Documents skipped because nothing changed are not re-read and so cost nothing. See Usage and plan.