Managing Documents in eLLM

What this product is for

For eLLM to answer from your company's knowledge, that knowledge needs to be in its document library. Uploading a file makes its content searchable, so when someone asks a relevant question the assistant can find the right passages and cite them. The documents page gives you control over what is in the library and who can use it.

Main features

• Upload files by drag-and-drop or by browsing, including several at once.
• Add a source straight from a web link, such as a web page, an online reference, or a published file.
• Assign access tags so only the right groups can see a document.
• Set an age-out date so a document is automatically retired later.
• Smart versioning that skips duplicates and replaces updated files.
• A searchable, filterable library list with per-document actions.
• Bulk delete, per-row download, and the ability to re-enable retired documents.
• An image library, upload photos and diagrams that eLLM can display in chat when asked.
• A video library, upload clips that eLLM can play inline in chat when asked.
• Organised into clear tabs so the page stays easy to navigate.

The page is arranged into tabs: Library (the document repository you can search and manage), Add files (uploading from your device and, for administrators, the server watch folder), Web sources (adding and managing pages added from a link), Images (an image library for displaying photos and diagrams in chat), and Videos (a video library for playing clips in chat). The tabs you see depend on your permissions.

How to use it

Uploading documents

• Open the documents page.
• Drag files onto the upload area, or select browse to pick them.
• You can select several files in one upload.
• If your account allows it, choose one or more access tags and, optionally, an age-out date.
• Start the upload. A progress indicator shows each file being processed, and the list updates when indexing finishes.

Supported formats include PDF, Word documents, and plain text. PDFs are processed page by page, and any diagrams or images on a page are captured so they can appear in citation previews.

Controlling who can see a document (access tags)

Access tags decide which groups can retrieve and cite a document in chat. When you upload, you can apply one or more tags if your account has the permission to manage them.

• A document with no tag is private to the person who uploaded it.
• A document tagged for a group can be used by members of that group.
• Visibility in the library list is filtered the same way: administrators see everything, while others see public documents plus those for their own groups.

Setting an age-out date

An age-out date is the point after which a document is automatically retired from answers. This is useful for time-limited material. You can set it during upload and edit it later from the library list. Retired documents are hidden by default but can be brought back.

How versioning works

ELLM handles repeat uploads sensibly so your library does not fill with duplicates:

• Identical file: if you upload a file that is exactly the same as one already indexed, the upload is skipped.
• Changed file, same name: the new version supersedes the old one. The old version is kept but marked SUPERSEDED and excluded from answers.
• Renamed file: for single-file uploads, a New version of picker lets you explicitly mark which existing document the new file replaces.

Superseded versions are never silently lost; they are hidden from answers but can be restored.

Finding and managing documents in the library

The library list is paginated and searchable. You can:

• Search by document name or by who uploaded it.
• Filter by department.
• Turn on Show hidden to include aged-out and superseded documents.
• Download any document from its row.
• Edit a document's age-out date inline.
• Re-enable a retired document.
• Select several documents and delete them in one action.

Bulk ingestion from a watch folder (administrators)

Administrators can trigger ingestion of files placed in a designated watch directory on the server, using the directory ingestion button in the Add files tab. This is handy for an initial migration from a network file share.

Adding a source from a web link (Add from URL)

If your group has the ingest web URLs permission, the Web sources tab lets you add a source by pasting a link instead of uploading a file. This is ideal for an online reference page, a published specification, or a file hosted on the web.

1. Open the Web sources tab.
2. Paste the link into Add from URL.
3. If your account allows it, choose access tags and, optionally, an age-out date, just as for an upload.
4. Start the import. eLLM fetches the page on the server, extracts the readable content, and indexes it like any other document.

A web page is turned into clean, readable text; structured data is kept tidy; and a document such as a PDF served at a link is processed just like an uploaded file. Imported web sources appear in their own list in the Web sources tab, where you can re-import one to refresh its content or remove it. Re-importing the same link updates that source rather than creating a duplicate.

This permission is separate from the organisation-wide internet search used in chat, so a chosen group can be allowed to add web sources without web search being switched on for everyone. For safety, eLLM only fetches genuine web addresses and refuses internal-only network addresses.

Using the image library

The Images tab holds a library of photos, diagrams, and other images that eLLM can show in chat. When the Documents toggle is on in chat and someone asks "show me an image of…" or similar, eLLM searches the library and displays the best match alongside its caption and source. If nothing relevant is found locally and the internet toggle is on, eLLM will show a cited web image as a fallback.

To upload an image:

1. Open the Images tab.
2. Use the upload card to select a JPG or PNG file.
3. Enter a caption describing what the image shows. The caption is what eLLM searches, so a clear description of the subject produces better matches.
4. Optionally add tags and set an age-out date, then upload.

The tab shows a thumbnail grid of all uploaded images. You can search by name, caption, or tags, and use the Edit control on any image to update its caption, tags, departments, or age-out date. Editing the caption re-indexes the image so search results stay accurate.

Using the video library

The Videos tab works in the same way as the image library but for video clips. When someone asks "show me a video of…" or "play a video of…", ELLM searches the library by caption and tags, and plays the best match inline in the chat window. Supported formats are MP4, WebM, OGG, and MOV.

To upload a video:

1. Open the Videos tab.
2. Use the upload card to select a video file.
3. Enter a caption and any relevant tags, then upload.
4. ELLM extracts a poster frame automatically so the thumbnail grid shows a preview.

Search, pagination, editing, and age-out work the same as for images.

Common tasks

• Add a new policy document: upload it, tag it for the groups who should see it, and let indexing finish.
• Add an online reference: open the Web sources tab and paste its link into Add from URL.
• Refresh a web source: re-import it from the Web sources tab to pull in the latest content.
• Replace an outdated document: upload the updated file with the same name to supersede the old one, or use the New version of picker for a renamed file.
• Retire something temporarily: set an age-out date, or delete it; re-enable later if needed.
• Find an old version: turn on Show hidden to reveal superseded and aged-out documents.
• Add a photo or diagram for chat: open the Images tab, upload the file with a clear caption.
• Add a training or reference video: open the Videos tab, upload the file with a caption and tags.
• Update an image's description: open the Images tab, find the image, and use Edit to update the caption eLLM re-indexes it automatically.

Things to know

• Uploading and tagging may be limited to certain accounts. If you cannot upload or set tags, your group may not have those permissions.
• A document with no access tag is private to its uploader.
• Superseded and aged-out documents are excluded from answers but are kept and can be restored.
• There is a maximum size per upload, set by your administrator. Very large files may be rejected.
• Indexing happens in the background; an answer can only use a document once indexing has finished.
• Image captions are what eLLM searches when someone asks to see a picture of something. A vague caption produces poor matches; a specific description of the subject produces good ones.
• Video captions and tags are the only searchable text for videos — there is no automatic transcription.
• Images and videos shown in chat are access-controlled the same way as documents: they are only returned to users in the groups the item is tagged for.

Troubleshooting

• My upload was skipped: an identical copy is already indexed, so nothing needed to change.
• I can't add access tags: managing tags requires a permission your group may not have. Ask your administrator.
• My file was rejected for being too large: it exceeds the upload size limit. Ask your administrator about the limit or split the file.
• A document I expected isn't in answers: check it isn't superseded or aged out (use Show hidden), and that it is tagged for a group you belong to.
• The upload spinner seems stuck: wait for indexing to finish; the list updates when it completes. If it fails, the status will show the failure.
• An image isn't appearing when I ask for it in chat: check the caption accurately describes the subject. Also confirm the Documents toggle is on in the chat toolbar, and that you are in a group the image is tagged for.
• A video won't play: the video must be in a browser-supported format (MP4 is the most reliable). If the format is correct, check the Documents toggle is on and that you are in the right group.

Frequently asked questions

What file types can I upload?

Common formats including PDF, Word, and plain text. PDFs keep their page structure and capture diagrams for citation previews.

What happens if I upload the same file twice?

If it is identical, the upload is skipped. If it changed, the new version supersedes the old one.

Can I get back a document I replaced?

Yes. Superseded versions are kept and can be restored. Turn on Show hidden to find them.

Who can see what I upload?

That depends on the access tags. With no tag, it is private to you. With a group tag, members of that group can use it.

Can I add a web page instead of a file?

Yes, if your group has the ingest web URLs permission. Use the Web sources tab and paste the link. The same access tags and versioning apply.

Do I need internet search switched on to add a web source?

No. Adding web sources is controlled by its own permission, separate from the organisation-wide internet search used in chat.

Does deleting a document remove it permanently?

Deleting removes it from the library. If you only want to retire it temporarily, set an age-out date instead, then re-enable it later.

Can eLLM show an image even when I haven't asked for one?

Yes. When the Documents toggle is on, a relevant library image may appear alongside a normal document answer if it matches the subject of your question. This is display-only; the image caption does not change the written answer.

Why doesn't eLLM transcribe video audio to make it searchable?

On-prem transcription requires additional software and processing capacity that is not included in the base installation. Search is based on caption and tags only. If searchable transcripts are important, ask your administrator.

Summary

The documents page turns your files and media into searchable knowledge for the assistant. Upload documents, tag them for the right groups, and let versioning keep things clean. Use the Images and Videos tabs to build a media library that the assistant can draw on in chat with a clear caption on each item, eLLM can find and display the right picture or clip when someone asks for it.