Adding and editing documents · Workforce Intelligence docs

The Documents card on a Knowledge Base page shows every document, with its Filename, when it was Uploaded, and its Status. The card is in the left column of the page, and its header shows the document count along with a drag-and-drop zone for uploading files.

Status	Meaning
Waiting	The file is stored and is queued to be read
Reading…	The file is being read. A note beside it says what is happening, for example which pages are being read
Available	The document is indexed and Coworkers can use it
Failed	The document could not be indexed. The reason is shown under the status, with Try again for people who manage the Knowledge Base (see Try again and Read again)

When a Knowledge Base has no documents, the list says none have been uploaded. You can drag files directly onto the drop zone in the card header, or use Upload documents and Create document manually further down the card.

Who can change documents#

Everyone who can see a Knowledge Base can open its documents, read them and download the original files. You can add, replace, edit and delete documents if you are the Knowledge Base's owner, or one of its editors. Other people see the documents read-only, with no upload controls, no Replace file, no Save and no Delete.

Supported file types#

Kind	File types
Text	`.txt`, `.md`
Word processing	`.docx`, `.odt`
PDF	`.pdf`
Spreadsheets and tables	`.xlsx`, `.ods`, `.csv`, `.tsv`
Presentations	`.pptx`, `.odp`
Images	`.png`, `.jpg`, `.jpeg`, `.webp`, `.gif`

The upload area lists the supported types under Drag files here, or browse. A file in another format is rejected with a message that names the file. Older Microsoft Office files (.doc, .xls, .ppt) are not read: the message tells you to save the file as .docx, .xlsx or .pptx and upload it again. Password-protected and encrypted files cannot be read either, and show Failed with the reason. A file larger than the size your administrator allows is refused with a message that states the limit; the upload stops as soon as the file passes it. A supported file whose contents are damaged is refused with "This file is damaged and could not be opened." Every refusal in an upload list starts with the file name, so you can tell which file it was.

Files that web sources find linked from a page are read the same way.

Add a document#

The Documents card gives you two ways to add content.

Upload files#

Drag one or more files onto the drop zone in the card header, or scroll to Upload documents and drag files onto the box there, or select browse and choose them. You can add many files at once.
Each file shows its own line: Waiting to upload, then Uploading…, then "added — reading it now". If a file cannot be added, its line says why.
Reading happens in the background. You can leave the page; the files keep being read and their Status changes from Waiting to Reading… to Available. While any document is still being read, the list refreshes by itself.

If you upload a file with the same name as a document already in the Knowledge Base, the new file replaces it (see Replace a file). If the file is identical to the one already indexed, the line reads "is already up to date" and nothing is read again.

Create a document manually#

Select Create document manually.
Enter a name in the filename.md field. If you leave it blank the document is named untitled.md.
Paste or write the text in the large field. The button stays unavailable until there is some content.
Select Create document, or Cancel.

A message reads "Created" followed by the file name and the chunk count.

What the Knowledge Base keeps#

A Knowledge Base is an index of your files. It stores the text it reads from each file so Coworkers can search and cite it, and it keeps the original file so you can get it back.

A document made from an uploaded file or a synced web page is read-only text. Its details say "This is an index of" the file name. Owners and editors also see how to change it: replace the file, or let the source update it.
A document you pasted or wrote by hand is editable text.

Open a document#

Select a document's file name in the list. Its details open below the row and show its size and, for files with pages, the page count.

Until a document has been read, the details show a short message in place of its text: that it is queued for indexing, that it is being read now, or, for a document that could not be read, the reason. A document with no readable text says so. The text appears once indexing finishes.

Button	What it does
Replace file	Uploaded files only; owners and editors. Swap in a new version of the file
Download original	Saves the file you uploaded
Open source	Web pages, linked files and SharePoint files only. Opens the original address in a new tab
Save	Pasted documents only. Saves your edits
Close (Cancel for a pasted document)	Closes the details

Edit a pasted document#

Change the text.
Select Save. A message reports how many parts were re-indexed and how many were reused, and the editor closes. Only the changed parts of the document are processed again.
Select Cancel to close without saving.

Replace a file#

Open the document and select Replace file.
Choose a new file of the same type, for example a new .pdf for a .pdf. A file of a different type is not accepted; the message suggests uploading it as a new document instead.
The document shows Reading… while the new version is read. Until it is ready, the previous version stays in use. When it is Available, Coworkers see the new text.

If the new version cannot be read, the document stays Available with the earlier text and a note, "The new version of this file could not be used: reason. The previous version is still in use." Select Try again on the document to read the file again (see Try again and Read again).

You can also replace a file by uploading a file with the same name.

Notes on how a file was read#

When something in a file could not be read in full, an available document shows Indexed with notes (N), where N is the number of notes. Select it to see them. A document that failed shows Notes (N) instead, with the first note written out under the reason. Typical notes are:

Some pages or pictures were not read, or only partly read, because they exceed a limit set by your administrator.
A spreadsheet or table was cut after the first rows allowed, or a document was cut after the allowed amount of text.
Hidden sheets of a spreadsheet were skipped.
Spreadsheet formulas had no calculated value and were left blank. Opening and saving the file in Excel calculates them.
Charts, diagrams or pictures were not read.
Only the first frame of an animated image was read.
A PDF with more pages than your administrator allows: the first pages were read and the note says how many.
Scanned pages or pictures were read with a different model, because the image-reading model your administrator chose is not available.

Everything else in the file is still searchable.

Scanned pages and pictures#

PDFs, Word documents and presentations often hold text that you can already select. That text is read directly and exactly.

Pages that are scans, standalone images, and large pictures inside Word, PowerPoint and OpenDocument files hold no text to read directly. These are read by an image-reading model when your administrator has chosen one. A document that was read this way shows Read from page images under its status. The text it produces is a transcription and can contain mistakes, especially with handwriting, low-resolution scans and dense tables. Small decorative pictures such as icons and logos are skipped.

If your administrator has not chosen an image-reading model, another model that can read images is used when there is one. If they chose one and it is not available, another model is used and the document carries the note "Read with a different model because the chosen image-reading model is not available." When no model can read images, text pages are still indexed, and the scanned pages or pictures are listed in the notes as not read. A file that has only scanned pages or only pictures shows Failed, with a message asking you to have an administrator choose an image-reading model.

Why a file could not be read#

The reason is shown under Failed. The common ones:

"This file took longer than N minutes to read. Try splitting it or asking an administrator to raise the reading time limit." Large presentations and scans with many pictures are the usual cause. Split the file, or ask an administrator to raise the reading time.
"This file is too large or complex to be read." Split the file into smaller ones.
A file over the size your administrator allows says so.
"The model that reads page images could not be reached." (or took too long, or was too busy). Select Try again later; if it keeps happening, tell an administrator.
A file that is damaged, password protected, or in a format that is not read says so in the same way.

When only some pages or pictures are over the limit your administrator set, the document is still indexed and a note names the limit, for example "Pages 301-420 were not read because they exceed the limit of 300 pages set by your administrator."

Try again and Read again#

People who manage the Knowledge Base (the owner and editors) can have a file read again from its original. Which button you see depends on the document:

Try again appears in the row of a document that shows Failed, and of one whose replacement could not be used.
Read again appears for an Available document that has a file behind it. It is in the row when the document has notes, a message under its status, or was read from page images, and it is always next to Replace file and Download original when you open the document. A document that is Waiting or Reading… shows neither.

Choose Read again when something has changed since the file was first read: your administrator chose a different image-reading model or raised a limit (see below), or pages or pictures were left unread. A confirmation asks "Read this file again?" and explains that it reads the original file again. When the document was read from page images, or its notes mention pages or pictures that were not read, it also says that scanned pages and pictures are read by the image model again, and that this is billed like the first time. Select Read again to start, or Cancel. A message confirms "filename is being read again."

Coworkers and search keep using the current version until the new one is ready. Then the new text, notes and Read from page images label replace the old ones together. If the new version cannot be used, the document keeps the earlier text and shows "The new version of this file could not be used: reason. The previous version is still in use."
The image-reading model and the limits in force when the read starts are the ones used. A change your administrator made after the file was first read does not touch it until someone reads it again, so an old note, such as one about a fallback model, stays until then.
For an uploaded file, the file is read again from the copy stored at upload, so you do not need to upload it again. The document shows Waiting and then Reading… as usual.
For a file a web source collected, the file is fetched again from its original address.
For a web page, the message tells you to sync the source instead.
A document with no file behind it, such as one you wrote by hand, has nothing to read again.

A document that stays Waiting or Reading… for a very long time is marked Failed with "Reading this file was interrupted. Try again." If reading could not be started at all, the message is "Reading this file could not be started. Try again."

How tables appear#

Spreadsheets, CSV and TSV files become tables, one per sheet. Each row is kept on its own line and each sheet is introduced by its name, so search finds the individual rows. When a long table is split for search, the header row is repeated with each part so a row always shows what its columns mean. Formulas show their last calculated value.

Delete a document#

Select Delete on the document's row.
In the confirmation, select Delete document. The dialog states that the document's indexed content is removed too, and that this cannot be undone.

You can delete a document collected from a web page the same way. If its web source syncs again, the page is collected afresh as a new document.

Rebuild search index#

If searches seem to miss passages that are in your documents, select Rebuild search index in the Documents card header. It is available to the same people who can change documents. On phones, the button shows only the refresh icon.

Select Rebuild search index. A Rebuild search index dialog says how many documents will be re-read and how many passages will be refreshed. If nothing needs changing, it says all documents are already up to date.
Select Rebuild search index to start. Keep the page open until it finishes.
A summary reports how many documents were processed and how many passages were refreshed. Any document that could not be read is listed by name; you can run it again.

The rebuild re-reads the text already stored for each document and splits it into passages again, so you do not need to upload anything. Files are not read again, and scanned pages and pictures are not sent to the image-reading model. Passages that did not change are kept as they are. Document titles and dates are not changed. Only one rebuild can run for a Knowledge Base at a time.

Where documents appear#

Coworkers cite documents in their replies. In a chat, you can open a cited document in the context pane.