Document Archive

AI ingest · OCR · 1C matching

checking…
Loading…

1C link: /documents?deal_id=<1C_DEAL_ID>

Process the scanned archive

Queues scans from the source folders into the intake tree; the worker picks them up and runs OCR, extraction and 1C matching. Originals are never modified — the folders are mounted read-only.

Loading…

Add a record by hand

For a document that exists in 1C but whose scan is missing or arrives later. It is stored with origin=manual so it can never be mistaken for something read off a scan.

Upload a scan

Runs the full pipeline: OCR, extraction, matching, archiving.

New request

A требование from the ФНС, a bank's 115-ФЗ query, an auditor's or a counterparty's letter. The scan is OCR'd, read into an itemised checklist, and every item is matched against the archive.

Open requests

Reindex

Replays extraction and 1C matching over documents already in the archive — no re-scan, no OCR, no AI spend. This is how a corrected document_types.yaml or a freshly synced deal table reaches older records. Fields a reviewer edited by hand are never overwritten.

Deleted records

Deletes are reversible by default: the row is hidden and the audit trail kept. A purge also destroys the stored objects and cannot be undone.