Imports and review

Import PDFs, images and documents

Biosphere reads invoices, price sheets, catalogs and photos page by page and suggests supplier details and product lines. You map and approve them like rows from a spreadsheet.

Know what Biosphere reads from a document

Each page is read for two kinds of details:

• Supplier details (supplier_fact): name, email, phone, line1, line2, city, region, postal_code and country.
• Catalog lines (catalog_fact): sku, title, description, supplier_name, price, currency, unit and availability_text.

Each detail keeps the short passage it came from, which you can see in the row’s details. Reading looks only for supplier and catalog details: it doesn’t create orders, receipts, payments or stock changes, and it doesn’t turn invoice totals into prices.

PDF, PNG, JPEG, DOCX, DOC, PPTX, PPT, RTF and plain text files are read this way. Each image or text file counts as one page. Spreadsheets are read as rows instead; see Import a CSV or spreadsheet.

Import a document

1. Upload the file in Inbox or on the Files page, or email it to your business address. See Upload files.
2. While pages are read, the source shows Processing. Open it to follow progress, such as “Extraction 1 · Extracting · 3 of 8 pages processed”, with each page marked Processed or Awaiting successful extraction.
3. When every page has been read, the source changes to Needs mapping and the details appear as rows. Nothing appears until every page succeeds.
4. In From, choose who issued the document, such as Alder Supply.
5. If both kinds of details were found, a menu at the top of the review lists catalog_fact and supplier_fact with their counts. Choose the one to work on.
6. Map it. For catalog lines, choose Catalog variants as the Record type and sku as the identifying field. See Map a source.
7. Select the Review changes button, check the changes, then select the Approve import button.

Approve only the details you trust

Biosphere marks details it was unsure of: “This fact contains uncertain values. Review the original page before mapping it.” Those rows show Needs attention and can’t be approved.

1. Compare them with the original using the Download original button.
2. Select Choose records and tick the rows to keep.
3. Select the Review changes button, then the Approve import button.

Check page coverage

Before the details are ready, coverage shows in the review. Afterwards, open Processing details to see it. For emailed documents, coverage also shows under Delivery health on the Email page.

Coverage shows the extraction version and state (Queued, Extracting, Ready for review, Needs attention or Cancelled), how many pages are processed, each page’s result, recent attempts and the version history.

Retry failed pages

Biosphere retries a failed page on its own first. The source shows when the next try is scheduled and which attempt it is.

If pages still fail, the coverage reads “Processing stopped. Your original and completed pages are retained.”

1. Use the Retry failed pages button (Retry failed slides for a presentation). Only the failed pages are read again.
2. If the retry limit was reached, read the whole document again instead.

Use the Inspect processing details link for the full attempt history on the Background jobs page.

Read a document again

1. Open the source in Inbox. If its details are already in review, open Processing details to find the page coverage.
2. Open Run OCR again.
3. Select Start new extraction.

Biosphere reads every page again from the original and creates a new extraction version. It may use paid cloud processing if fallback is enabled. Earlier versions and their results stay. You can’t start one while another extraction is running.

What to know

• Each file can be up to 25 MiB and 100 pages or slides. Images can be up to 25 megapixels.
• Locked PDFs, and PDFs with scripts, active content or embedded files, are refused. Upload an unlocked copy without them.
• DOCX, DOC, PPTX, PPT and RTF files are turned into pages before reading. The original stays as it was uploaded.
• Approving document details goes through the same review as any other import. Nothing changes until you approve.