Importing files
Import adds files you already have, such as policies, handbooks, spreadsheets and slide decks, to the wiki in one go. By default, files are added exactly as they are: the originals are kept. If your workspace has a Librarian, you can also ask it to turn some files into wiki pages; see Turning files into pages. Where a file contains text, 3mote reads it so it can be searched, and so the Librarian can answer questions from it.
Starting an import
There are several ways in, and they all open the same Import documents dialog:
- Click Import on the Wiki page, or in the header of the Librarian chat.
- Drag files anywhere onto the Wiki. A panel reads Drop to import into the wiki; let go and the dialog opens with your files listed and the folder you’re viewing already chosen. Dropping onto a folder in the rail opens the dialog with that folder chosen. While you’re editing a page, a dropped file is embedded in the page instead. You can also drop more files onto the open dialog. If your workspace has a Librarian, its home also has a drop area that says Drop documents here to add them to the wiki.
- Click + New at the top of the wiki rail and choose Import files….
You see these only if you’re allowed to add to the wiki; see Who can import below.
In the dialog, drag files in or click Choose files. Add more adds to the list. Picking or dropping the same file twice adds it once.
What you can import
- Types: PDF, Word (.doc, .docx), PowerPoint (.ppt, .pptx), Excel (.xls, .xlsx), CSV, text and Markdown (.txt, .md), JSON, images (JPG, PNG, GIF, WebP) and video (MP4, WebM, MOV).
- How many: up to 50 files per import. If you choose more, the extra files are left out and you’re told.
- How big: videos up to 200 MB, every other file up to 100 MB.
- What can be read for search: PDFs with selectable text, Word (.docx), PowerPoint (.pptx), Excel (.xlsx), and text, Markdown, CSV and JSON. Older Office formats (.doc, .ppt, .xls) can be imported but not read; save them in the newer format first.
Files count toward your workspace’s storage; see Plans and storage.
Choosing where files go
Add to sets the destination: Top level, any existing folder, or New folder…. New folder… appears only if you can create folders (it reads New folder in name … when you’re inside a folder). When nothing else decides where files go, the dialog chooses the folder you last imported into. Choose New folder…, type a Folder name, and 3mote creates it as the import starts (you’ll see Creating folder…). When the import finishes, the destination folder opens in the rail.
Naming your files
Each file gets a suggested name, cleaned up from its filename. Underscores and hyphens become spaces, the extension is dropped, and leftovers such as “final”, “copy” or “(1)” are removed, so Harborlight_White_Paper_FINAL_v3 (1).pdf becomes “Harborlight White Paper”. A name written all in capitals or all in lower case is also capitalised like a sentence. The cleanup only removes, re-spaces or capitalises what is already in the filename; it never writes a new name for you.
You can edit any name before importing, and a name can’t be left empty. The original filename is shown under it as Original file while you import. Afterwards the file is listed, and downloaded (with its file extension), under the name you choose.
If a name is still hard to recognise, such as a camera or scanner name, you’ll see This name is hard to recognise. Give it a clearer one? It’s only a hint; you can import anyway.
While it imports
Click the import button (Import 12 files, for example). Each file shows its progress, then Uploaded. Finishing… while 3mote reads its text, then Added. Files are processed in order, and a screen reader is told how many have finished.
When every file is done, a summary lists how many were added and where, and how many of them are ready to search. It offers Import more, Done, and Ask the Librarian if your workspace has one.
What each result means
Under each file you’ll see one of:
- Ready — the Librarian can answer from this — its text was read. It is searchable, and the Librarian can use it. This needs a Librarian to be set up. Without one, the line reads Ready — searchable in the wiki.
- Added, but not searchable: older Office formats can’t be read. Save it as .docx, .xlsx or .pptx to make it searchable. — .doc, .ppt and .xls files are added as they are.
- Added, but not searchable: no text found (it looks like a scanned document) — the file was added, but it holds only pictures of text. Scanned documents can’t be read yet.
- Added, but not searchable: images aren’t searchable yet — images are added but their contents aren’t read.
- Added, but not searchable: this file type isn’t searchable — for example video.
- Added, but not searchable: files over 10 MB aren’t read yet. — the file was added but is too big to be read for search. A very long document may also be read only in part. It still opens and downloads normally.
- Added, but it could not be indexed for search — something went wrong while reading the text. The file itself is safe.
A file that is added but not searchable is still in the wiki: you can open, download, share and comment on it as normal. See Search and action items.
When something fails
A file 3mote won’t accept, such as an unsupported type or an empty file, shows Not added. It can’t be retried; choose a different file. Other failures show the reason and a Retry button. Retry failed retries every failed file at once. Files that already succeeded aren’t uploaded again.
Cancelling and leaving
Cancel a single file with the X on its row. If its upload is partway through, 3mote asks Cancel this upload?. Cancelling stops the transfer and removes the file. Keep uploading carries on.
Closing the dialog while files are uploading asks the same question. If you’ve chosen several files, or renamed one, and haven’t imported them yet, closing asks Discard these files?. Keep editing returns to the dialog.
Leaving the Wiki for another part of 3mote while files are uploading asks first; moving around inside the Wiki doesn’t. Files that have already finished stay in the wiki, and a file that is still uploading shows a progress indicator in the rail with no row menu until it finishes.
Turning files into pages
If your workspace has a Librarian that is ready, the dialog offers a choice under How to bring these in: the Librarian writes a page from the file’s text, names it from what is in it, and attaches the original. You choose per file type, or per file. Without a Librarian, or while it is unavailable, the dialog tells you that files will be added as they are, and everything works as described above. Only files you uploaded yourself can be turned into pages.
Choosing
For each group of files, such as 1 PDF or 3 Word documents, pick Turn into a page or Keep as file. Images start as Keep as file. The other types that can become pages start as Turn into a page. Mixed shows when the files in a group differ. Click Choose for each file to decide one file at a time, and Hide the per-file choices to go back.
PDFs, Word (.docx), PowerPoint (.pptx), text, Markdown and images (PNG, JPEG, GIF, WebP) can become pages. Spreadsheets, data files such as CSV and JSON, and other types always stay files. Older Office files stay files too.
Review
When you click the import button and have chosen at least one page, a Review before the Librarian starts step lists every file. Each one reads Will become a page or Will stay a file, with the reason when 3mote already knows the file can’t become a page. You also see an estimate of how many tokens the Librarian will use to read and write. If the estimate is more than the Librarian has left today, you’re told the import may stop early and can be continued later.
People who can manage AI agents, and the agent’s accountable person, also see how many tokens are left today and the Estimated cost. Other people don’t.
If someone else’s import is running, the step says so with a line that begins Starts after and names that person’s import and how many of its files are done. Your import then waits in line instead of starting at once.
Click Start to begin, or Keep as files to add everything as it is. Nothing is sent to the AI provider before you click Start.
Following progress
The dialog shows Turning files into pages with a count of files finished. Each file moves through Waiting, Reading, Writing the page, Checking against the original and Filing, then Done, Kept as file or Skipped. You can close the dialog: the import keeps going, and the Import button shows a dot while it runs. Only one import runs at a time in a workspace. If someone else’s import is running when you click Start, yours waits in line; see Waiting in line. Each person can have one import running or in line at a time, and can start or continue an import 3 times an hour. Joining the line counts as a start.
Click Cancel import to stop. Files not yet done stay as files, and anything already created is kept.
Waiting in line
When another import is running, the dialog shows Your import is in line. It tells you how many imports are ahead of yours, and whose import is running and how many of its files are done. You never see the names of another person’s files. When it is your turn, your import starts by itself.
You can close the dialog. The Import button shows a dot while your import is in line, and you get a notification when your import starts. Click it to open the import.
- Cancel import takes your import out of the line. Your files stay in the wiki as files.
- You can have one import running or in line at a time. If you already have one, Start is disabled and a note says so until the first one finishes or is cancelled.
- The line can pause, and the dialog says why. If the Librarian has reached its usage limit for today, your import starts by itself when the limit resets, or as soon as an admin raises the limit and you open your import. If the Librarian is unavailable, your import carries on once that is fixed.
- A workspace that has reached its page limit or text limit does not pause the line. Each import stops with that reason instead, as described under When an import stops.
- An import that has been in line for 7 days is stopped, and you can continue it later.
What the Librarian writes
For each file, the Librarian reads the text in parts and writes a clean page with the same wording. Headings and lists are kept. Strikethrough and checklist ticks are not carried over: struck-through words appear as ordinary text, and a checklist becomes a plain list. It doesn’t summarise or add text. In PDFs and PowerPoint files of three or more pages, 3mote first removes page numbers (a line that is only a page number, such as 3, Page 3 or xii, and that follows the page sequence) and header or footer lines that repeat at the top or bottom of many pages. A header or footer that repeats word for word is kept where it first appears. This does not apply to Word, text or Markdown files. The page notes how many lines were removed. When the document’s first heading is the same as the page title, it isn’t repeated at the top of the page.
The Librarian names the page with the document’s own title, written out in full and in ordinary capitalisation, even when the document prints it in capitals (page names are limited to 120 characters). When the document’s first heading states the title more completely, that heading may become the name instead. If the document shows no title, the Librarian writes a short name for what it is, and if no usable name comes back, the page is named after the file. Because the title is the page name, the page’s first heading is taken out when it repeats the name, or at least its first three words, together with a subtitle heading right after it when the two make up the whole name. Pictures before it stay. Import details on the page quotes the heading that was taken out, so its wording is still there if you rename the page later.
Tables that are real text in a PDF are rebuilt as tables when their columns can be told apart. Otherwise their text is written row by row.
Pictures embedded in a PDF are brought into the page where they sat, including tables and charts that are pictures in the PDF. They count toward your workspace’s storage and follow the page’s access like any image in a page. Some pictures are left out on purpose: very small ones such as icons, a picture repeated on three or more pages such as a logo or letterhead, page backgrounds, drop shadows, glows and other soft single-colour layers, and very large pictures (over about 6 megapixels). A blurred greyscale picture, or a very pale one-colour shape, with no sharp edge anywhere can be mistaken for such a layer and left out. A document brings over at most 30 pictures and about 20 MB of pictures. Charts or drawings made of vector graphics, which are drawn rather than embedded as a picture, can’t be brought over yet. When such a drawing has a caption like ‘Figure 3’ or ‘Chart 2’, the page says which page of the file it was on; others are not reported. The original file is attached. If a picture’s place in the text is lost, the page is still created and notes how many pictures were left out. Pictures are brought over from PDFs only, not from Word or PowerPoint files. Reading pictures has time and size limits. If a limit is reached part-way, the page keeps the pictures read before it and says how many pictures, or pages of pictures, were not read. If the pictures can’t be placed when the page is created, the page says how many were left out. If a PDF’s pictures can’t be read at all, the page is made from its text only and says so. The page goes into the folder you chose, and the original file is attached to the page. Anyone who can open the page can open the original from it: View original at the top of the page opens the file: a PDF or picture is shown in a viewer, and any file can be downloaded from there. The page is published as soon as it is created.
Markdown files are different. They are copied as they are, named from their first heading, without the AI provider, and they get no percentage.
The uploaded file is removed when its page is created. The exception is when you aren’t allowed to delete your own pages, or the file was moved or replaced in the meantime. Then the file stays beside the page, and your workspace stores both the original and the copy attached to the page.
The Librarian appears as the author in the page’s version history. You own the page, and the page shows who imported it and from which file.
Results and names
Import results lists each file as Page created, Kept as file or Skipped, with Open page for new pages. For a new page you can click Accept to keep the suggested name, Edit to change it, or Use file name to name it after the file. The original file name is always shown.
A file that could not be turned into a page is marked as not transcribed. It shows a banner: This file could not be turned into a page, so it was kept as a file. with the reason.
The check
3mote compares each page made from text with the text of the original. Every number in the file’s text, apart from the page numbers, headers and footers removed above, must also appear in the page. Each part is checked and gets one retry, and then the finished page is checked again as a whole. The page shows Checked against the original with a percentage for the whole page. If either check fails, no page is made: the file stays a file and the reason is shown.
A page made from an image can’t be compared automatically, so it is marked Unverified transcription and shows a warning at the top: “This page was written from an image and has not been checked against it. Read it against the original before you rely on it.” Anyone who can edit an imported page can read it and click Mark as checked; the warning goes away and the page then reads Checked by a person.
The check is an aid, not a guarantee. Read important pages before you rely on them.
Files that stay files
A file stays a file, with the reason shown, when:
- it is a scan with no text to copy;
- it is password protected;
- it is too large to read: over 10 MB for documents, over 25 MB for images;
- the document is too long for one page (more than about 300,000 characters of text, roughly 100 to 150 pages);
- the page was too long for the Librarian to write in one go;
- the Librarian isn’t allowed to read it;
- you already imported the same text, either earlier in this import or as a page that still exists (images aren’t compared);
- the Librarian’s model can’t read images, or no readable text is found in an image; or
- the model declined or returned an error.
When an import stops
An import stops, and can be continued, when:
- the Librarian reaches its usage limit for today;
- the workspace reaches its page limit or text limit on the Free plan, or is out of storage;
- you lose access to the destination folder;
- the Librarian becomes unavailable: its key is rejected, it is deactivated, or the Wiki is switched off;
- the workspace’s Librarian is changed;
- it waits in line for 7 days without starting; or
- something goes wrong three times in a row, or the import makes no progress for a day.
Nothing already created is removed. A stopped import shows the note An import is waiting. Click Continue import to carry on once the cause is fixed, or Discard. A continued import goes to the back of the line. If files were left at the review step, the note reads Files are waiting for review: click Review to look again, or Keep as files.
Access, the AI provider and what is kept
The page gets the same access as the uploaded file. That follows the folder, unless you changed the file’s own sharing. A health-information marking on the file carries over to the page. The Librarian can read only files that both it and you can open: a file you have restricted is only turned into a page if you shared it with the Librarian first, and otherwise it stays a file. If your workspace marks content as protected health information, the Librarian doesn’t read those files unless the workspace allows agents to.
The Librarian uses the AI provider and model your workspace chose. To turn a file into a page, the file’s text, or the image, is sent to that provider. Markdown files are not sent. The usage counts toward the Librarian’s daily token budget and shows in its Runs tab as time, tokens and cost.
The text read from a file, and the partly written page, are deleted when that file is finished, kept as a file, cancelled or stopped. An import that stalls is stopped after a day, and anything left is deleted 7 days after the import ends. Files left at the review step are cancelled after 7 days. The import’s own records are deleted after 30 days. What stays with the page is the original file name, who imported it, the AI provider and model, the file’s type and size, a fingerprint (hash) of the text, and the check result. See the Privacy Policy.
Who can import
Importing follows the same permission as creating a page, so you can import if you can use New page. The same applies to turning files into pages. By default that is the Account Owner, Administrators, Managers and Members. Commenters, Readers and Guests can’t. If a role has been customised, check Roles and permissions.
Imported files respect wiki access like any other item: a file you import is visible according to the folder it’s in. See Sharing and access.
Related
- Wiki — the page tree and the row menu.
- The Librarian — asking questions of what you’ve imported.
- Search and action items — opening, downloading and deleting files.