Parse a received invoice or receipt (PDF / image / DOCX / scan) with the local docling tool into readable text - so any text model can read it - then extract vendor, number, amount, currency, and dates. Preview-first: shows the extracted fields and writes nothing; set write_mode to stage to persist a staged invoice/expense row into data.shared. Writes no accounting system - staging is the holding area for later review and posting. Content-isolated via Gateway-gated file refs.
Invoice intake organizer
Parse a received invoice or receipt (PDF / image / DOCX / scan) with the local docling tool into readable text - so any text model can read it - then extract vendor, number, amount, currency, and dates. Preview-first: shows the extracted fields and writes nothing; set write_mode to stage to persist a staged invoice/expense row into data.shared. Writes no accounting system - staging is the holding area for later review and posting. Content-isolated via Gateway-gated file refs.
Complete workflow
Vendor-neutral document-intake pack for a small business that keeps its own books. The two recipes parse received invoices/receipts and stage the extracted fields, with no accounting-vendor binding of their own: invoice-intake-organizer parses one received document with the docling toolkit and stages the extracted fields (preview-first); invoice-intake-watch polls a Google Drive folder for new documents and stages each through the same docling parse (download straight into the warehouse CAS as a file_ref, no bytes through op-step values). The accounting reporting workflows (overdue-invoice-chase, accounting-daily-back-office) moved to the accounting capability packs (accounting-quickbooks / accounting-xero) where they resolve their canonical op-steps against the bound vendor — install one of those for the AR / overdue / cash-in side. Bind the Google Drive connection once and assign each recipe a dish.
You are viewing one recipe in this workflow. Installing it opens the complete pack so its other recipes and background automations arrive together.
Watch a Google Drive folder for new invoice / receipt documents and stage their extracted finance data. Polls the folder via the gdrive file.list operation, picks the newest document that is not a Google-native file, downloads its bytes straight into the warehouse CAS as a Gateway-gated data.file (file.download surfaces result.file_ref, so the bytes never flow through op-step values), parses it to Markdown with the local docling tool (which materializes the file_ref to a temp file it reads), and pulls vendor, number, amount, currency, and dates from the parsed Markdown via the same file-ref isolation. Preview-first: by default it shows the extracted fields and writes nothing; set write_mode to stage to persist a durable staged invoice/expense row into data.shared for later review and posting. A per-run context guard skips re-processing the same newest file across polls. This is the Drive-folder source lane for the SMB-finance intake; the manual single-document lane is invoice-intake-organizer.
How it works
Inspect the data fetches, transforms, gates, and output this recipe runs.
Settings
Configurable at install. Defaults shown — change them anytime in Recued.
doc type
setting
=
invoice
write mode
setting
=
preview
source path
setting
=
staging prefix
setting
=
smb_finance_staging
confidence threshold
setting
=
0.6
Trust & control
What installing this recipe would let it do. Recued grants these permissions at install — review them there before approving.