Draw the fields once. Every matching document in the batch follows, and each cell still opens the page it came from.
Fund administration, audit, and controllership teams keep meeting the same file again: a capital call, a confirmation, a form. The fund name, the investor, and the amount sit in the same place every time. Someone still retypes them, and someone else still has to find that sentence in the PDF before they sign off.
Three things pile up together.
Take a capital call notice. On the first PDF you snip three regions: fund name, investor name, and total amount called. DataSniper names each snip, and each name becomes a column. Drag the snips into the order the sheet should use. Export writes one row for that notice and one row for every other notice you uploaded with it.
The boxes are page-relative. They read the same rectangle on page one of each file when the administrator’s template is stable. When a later notice inserts a line, DataSniper follows the label beside the first snip, marks that row, and the preview opens on the moved region. If a later notice wraps a long investor name past the box you drew, you open that cell, widen the snip on the preview, and the cell updates with the text now inside the box.
Seven synthetic notices in the sample set share that template. Notice 04 carries a trust name long enough to show the clip: a narrow snip stops at “The Johnathan Q. Abernathy Rev”, and a wider one returns the full “The Johnathan Q. Abernathy Revocable Living Trust dated March 14, 1998”. Notice 07 inserts one extra line so the fields move down.
Open the batch in the Excel task pane. Drag a rectangle on the first document and give it the field name you want as a header.
The same rectangles run across every uploaded file that shares the layout. You reorder the snips before export so the columns match the workpaper.
Click any exported cell. The source document opens with that snip highlighted. Drag or resize the box and the cell follows.
DataSniper keeps the extraction inside Excel. You bring the evidence in, you decide the fields, and the sheet keeps a path back to the page.
Upload the PDFs, or PNG, JPEG, and WebP images, and read them in the task pane. Drag across the fund name, the investor, the amount called, or whatever regions this template uses. Name the snip. That name is the column.
Each snip is an output column, with a Source document column beside them. Drag the snips in the list until the headers sit in the order you want to export. The reorder happens before anything is written to the sheet.
Export reads every uploaded document with those page-relative boxes. PDFs use the text layer inside the rectangle, so a box that stops mid-name returns only the characters inside it. You get one row per document in a table named DataSniperExport, starting at A1 of the active sheet.
With the task pane open, click a result. A preview shows that file and highlights the snip. If the investor name was cut off, widen the box on the preview. Extraction runs again and the Excel cell updates immediately. The edit applies to that document’s snip.
On a PDF with a text layer, DataSniper keeps the characters whose positions fall inside the box. The engine is PDF.js, loaded with the add-in and run on your machine. A narrow box truncates. It does not pull the rest of the line in from outside the rectangle.
Scanned pages and image uploads have no text layer. DataSniper then runs Tesseract.js in the browser on the cropped region. The first OCR pass downloads the library and the English traineddata from jsDelivr. The crop is recognized locally. If that download is blocked, the cell is left blank and the status line says why.
The rectangles only line up when the documents share a layout. A notice that moves the investor name will return whatever now sits in that box. Draw the snip on the template you actually received.
DataSniper is built for a batch that repeats. The examples below are the same action: mark the fields once, export a row per file, and review from the cell.
Work through a capital call series without retyping the notice.
Turn a stack of same-layout confirmations or notices into a sample sheet you can review.
Pull a recurring statement or notice into the close file when each copy shares the layout.
Extract the same boxes from a stack of one form.
Build the data table for a client pack that repeats the same schedule.
Give a reviewer the cell and the evidence together.
DataSniper is a task pane add-in. There is no DataSniper account and no DataSniper server that receives your files. Upload stores the bytes in IndexedDB for this site, inside the Excel webview on desktop or inside the browser for Excel on the web. The sheet stores the extracted text and a pointer back to that local file. Clearing site data removes the files. The cell text remains, and the preview can open a file only while that local copy is still there.
Desktop Excel and Excel on the web keep separate stores. A batch uploaded in one does not appear in the other. The add-in itself loads over HTTPS from this site, and Office.js loads from Microsoft. That traffic is the add-in. OCR, only when a page has no text layer, also fetches Tesseract.js and English traineddata from jsDelivr. The cropped image is not uploaded.
The browser harness is a stand-in workbook for trying the flow and for automated tests. It shows the table in the page. It does not write an .xlsx file.
PDFs are the primary format. PNG, JPEG, and WebP images of a page can be uploaded and viewed as well. A PDF with real text is read from its text layer. A scan or an image uses the OCR fallback described above.
You snip the first one. Those rectangles become the template for the rest of the batch. When a later notice inserts a line and a field moves, DataSniper follows the label beside the first snip, such as Investor or Amount called, and marks that row so you can check the shift. You can still adjust one document’s box from its preview, or apply that box to every document.
A table named DataSniperExport at cell A1 of the active sheet. The header row is Source document plus your snip names, in list order. Each uploaded document is one body row. An amount that is clearly currency, with a symbol, a thousands comma, parentheses, or a K, M, or B suffix, is stored as a number. The text from the page stays with the cell. The workbook also keeps the template and each value’s document, page, and region, so preview and a later export still work after you reopen the file. Export replaces an earlier DataSniperExport table and overwrites that range, including source links from the previous grid.
Click a result cell while the DataSniper task pane is open. Excel opens a dialog on that document with the region that produced the cell, including a label match that moved the box. After you move the box, Apply this box to all documents copies that rectangle onto the template and reads every loaded row again. A nudge that lands on a different field does not replace the label match. If the file is no longer stored in this browser, DataSniper asks you to choose that same filename. A workbook that already has DataSniper links opens the task pane again the next time that file opens.
DataSniper runs in Excel on Mac, Windows, and the web. DataSnipper’s desktop add-in runs in Excel on Windows, and DataSnipper also offers a separate Excel Online add-in. You sideload manifest.xml. The steps for the Mac wef folder, a Windows shared-folder catalog or Upload My Add-in, and Upload My Add-in on Excel for the web are below. DataSniper is distributed as this manifest. It is not an AppSource listing, and this page does not deploy it from an admin center.
The file bytes stay in local storage for this site. They are not posted to a DataSniper backend. The add-in pages and, on a scanned page, the OCR language data are downloaded. The recognition itself runs in the browser.
Sideload the manifest yourself. The task pane then loads from https://sheetspeed.agi-jarvis.com/datasniper/. Quit Excel completely before you copy a manifest into a folder. After it is installed, the Home tab shows a DataSniper button. If the group is missing, look under Insert, Add-ins, My Add-ins.
Network blocks sheetspeed.agi-jarvis.com? Use this alternate manifest and its starter workbook.
These steps follow Microsoft’s sideload path for Microsoft 365 Excel on Mac. Quit Excel from the Excel menu before copying the file.
DataSniper
~/Library/Containers/com.microsoft.Excel/Data/Documents/wef.wef is missing, create it. The parent Documents folder is there after Excel has been opened at least once.manifest.xml into wef. The file name can stay manifest.xml.Recent Microsoft 365 builds also show Upload My Add-in under Insert, Add-ins, My Add-ins. When that command is present, select the manifest from there. When it is absent, use the wef folder. Microsoft’s folder method is documented in Sideload an Office add-in on Mac.
Use the command your Excel shows.
An organization can hide Upload My Add-in. When you do not see it, use a shared-folder catalog.
C:\DataSniper, and put manifest.xml in it.\\localhost\DataSniper.Microsoft’s steps are in Create a network shared folder catalog.
Upload My Add-in on the web applies to your own account. It is a personal sideload, not an organization-wide deployment. If the command is missing, an admin has turned sideloading off. The manifest’s pages have to stay reachable at https://sheetspeed.agi-jarvis.com/datasniper/.
The notices share one template and differ in fund, investor, and amount called. They are synthetic. They are not demands for payment. Snip the three fields on notice 01, export, and then open notice 04’s investor cell if you want to see a narrow box truncate a long name. Notice 07 is the same call with one extra line, so a label match has to follow the fields.
| File | Fund | Investor | Amount called |
|---|---|---|---|
| capital-call-01.pdf | Northwind Growth Fund III, L.P. | Harborview Pension Trust | $1,250,000.00 |
| capital-call-02.pdf | Northwind Growth Fund III, L.P. | Cedar & Pine Family Office LLC | $875,400.50 |
| capital-call-03.pdf | Blackpine Opportunistic Credit Fund II | Meridian University Endowment | $2,100,000.00 |
| capital-call-04.pdf | Blackpine Opportunistic Credit Fund II | The Johnathan Q. Abernathy Revocable Living Trust dated March 14, 1998 | $640,250.75 |
| capital-call-05.pdf | Sable Ridge Venture Partners IV, L.P. | Oak Street Capital Advisors | $3,450,000.00 |
| capital-call-06.pdf | Sable Ridge Venture Partners IV, L.P. | Alexandra Catherine Pemberton Worthington III | $424,242.42 |
| capital-call-07.pdf | Northwind Growth Fund III, L.P. | Northline Municipal Pension Plan | $424,242.00 |
An image-only twin of the first notice is at capital-call-image.png. Coordinates measured from the text layer are in layout.json. You can also load notices 01–07 from the Load sample notices button inside the task pane.
Open the browser harness if you want to walk the snip, reorder, export, and preview flow before sideloading.
