Documents

Documents are the files your business already has: PDFs of brochures, a Word handbook, a price list, an exported FAQ. Upload them and your agent reads the text inside, so it can answer from the same material your team relies on.

6 min read Updated

You manage documents from the Knowledge Base of any agent in the dashboard. Open Knowledge Base → Documents to see the files this agent already knows, upload new ones, and control what visitors are allowed to download.

Your agent only knows the text it can read from the files you upload here, plus the other sources in your knowledge base. It cannot see anything you have not added.

Supported file types and limits

The upload area accepts seven document types. Pick whichever format your content already lives in. There is no benefit to converting a file first.

SettingWhat is allowed
File types.pdf, .doc, .docx, .rtf, .txt, .csv, and .md
Maximum per file10 MB
Maximum per upload100 MB in total, and up to 10 files at once (fewer if fewer knowledge slots are free)
How files countEach document uses one knowledge slot, shared with website links, FAQs, and your other sources

If you select files that do not fit, the upload area sorts them into clear groups so you can see exactly why each one was skipped: an unsupported type, a single file over 10 MB, or files that would push the batch over the 100 MB total. Files beyond the 10-file maximum, or beyond the knowledge slots you have left, are skipped with a note. Supported files in the same selection still upload normally. If you are already at your plan's knowledge-source limit, Add Documents offers an upgrade instead.

Spreadsheets and web pages. Excel files (.xlsx) are not supported. Save the sheet as .csv and upload that instead. Web pages (.html) do not belong here either. Add the page with Website Links, which strips the menu, header, footer, forms, and buttons out first and gives your agent far cleaner text than an HTML file would. For a page that is not public, save it as a PDF and upload that.

How to upload documents

On the Documents page, click Add Documents to open the upload window. From there you have two ways to add files.

  1. Drag and drop, or browse

    Drag files straight onto the dashed upload area, or click Select Files to pick them from your computer. You can add several documents in one go.

  2. Watch each file upload

    Every file shows a spinner while it transfers and a check mark once it is in. Anything that does not meet the rules above is listed separately with the reason, and you can remove a file from the list before continuing.

  3. Confirm the upload

    Click Upload. The window closes and your new documents appear in the list while the agent reads them. You can keep working in the meantime.

The Documents page of the knowledge base showing uploaded files with status icons and download toggles
The Documents page. Each file shows its status, an upload date, a view button, and a download toggle.

What happens after you upload

Adding documents to your agent is automatic. For each file, Ultimo Bots:

  • Reads the text out of the file. It pulls plain text from PDFs, text, CSV, and Markdown files, and paragraphs plus table rows from Word documents, so pricing and spec tables come through too.
  • Keeps CSV rows apart. A CSV file that looks like a table (at least two rows and two columns, with most rows the same width) is turned into one record per row, each labeled with its column names, for example Product: X, Price: Y. A row is never merged with its neighbours, so the agent finds the right line of a price list or product table. The file still counts as one knowledge source.
  • Adds it to the knowledge base. The extracted text is split into smaller passages and indexed so your agent can find the right part when a visitor asks a question.
  • Answers from the extracted text. Your agent works from the text it read, not from the file itself. The original is kept on file so it can be offered as a download when you turn on the download link for that document.

Each document moves through a short lifecycle, shown by the icon next to its name in the list:

StatusWhat it means
ProcessingThe file is being read and indexed. A spinner shows on the row, and the row is locked until it finishes.
ReadyA check mark means the document is indexed and your agent can answer from it.
FailedThe file could not be read, for example because it has no readable text or is larger than 10 MB. The reason is shown under the row. Fix the file, remove the failed entry, and upload it again. A failed document does not count toward your limit.
Check what your agent actually read. Click the view button (the eye icon) on any document to see the exact text that was extracted. This is the fastest way to confirm a PDF produced real text rather than an empty page.

Share a download link in chat

By default, a document is used only as knowledge: your agent reads it, but visitors cannot get the file. If you want visitors to be able to download the original, for example a brochure or a price sheet, turn on the download link for that document. When it is on, your agent can include a download link for that file in its chat replies, so a visitor can open and save the full document.

  • One file at a time. Use the Share download link in chatswitch on a document row to turn the link on or off for just that file.
  • All at once. Use the Share download links in chat switch at the top of the list to flip every document together.
Only share what you mean to. Internal handbooks, contracts, or notes make good knowledge but are usually not meant for visitors. Leave downloads off for anything private. The agent can still answer from a file with downloads turned off.

Remove a document

To remove a file, select it with its checkbox and click the trash icon, then confirm. You can also select several documents and remove them together. Removing a document deletes it from the knowledge base and from the search index, so your agent stops using it right away.

There is no in-place edit for a document. If a file changes, upload the new version and remove the old one so the agent does not answer from outdated content.

Good to know

Scanned or image-only PDFs have no text to read. If a PDF is a picture of a page rather than real text, there is nothing to extract and the document will end up empty or failed. Upload a version with selectable text, or paste the content into a text or Word file instead.
Password-protected files cannot be opened. Remove the password from a PDF or Word document before uploading it, otherwise the text cannot be read.
Your agent only knows what you give it. It answers from the documents and other sources in this knowledge base. If a detail is not in any of them, the agent will not be able to provide it, so upload the material that covers the questions you expect.

Next steps

Documents are one source of knowledge. Round out what your agent knows: