Skip to content

Files, YouTube & Text Sources

Train your bot on content that isn't a webpage — PDF and text file uploads, plus the YouTube and Text source types you'll see in your source list.

On this page

Not all knowledge lives on a webpage. This article covers training your bot on uploaded files, and explains the YouTube and Text source types you may see in your source list.

Uploading files

You can upload files from your bot's Sources tab:

  1. Next to the Training URL input, open the Upload menu.
  2. Choose PDF or text files.
  3. Pick one or more .pdf or .txt files. They appear in the pending list below with a File badge.
  4. Click Start training.

(The other Upload option, CSV of URLs, is for adding webpages in bulk — see Training from a Website or Sitemap.)

Supported file types

Type

Extensions

Notes

PDF

.pdf

Text is extracted from the document — see the note on scanned PDFs below

Plain text

.txt

Ingested as-is

How files are processed

Files are processed as soon as you start training — there's nothing to fetch, so they usually finish in seconds. HelpJet extracts the text, splits it into chunks, and stores the embeddings in your bot's knowledge base. Processed files show in the source list with their size and page count.

What files work well

  • Product manuals and spec sheets
  • Policy documents (returns, warranties, SLAs)
  • Internal FAQ documents that never made it onto your website
  • Onboarding or how-to guides

Skip heavily visual documents — slide decks and image-rich brochures extract poorly, because only the text survives.

YouTube sources

YouTube is a source type in HelpJet — you may see a YouTube filter tab and video entries (with their duration) in your source list. Adding new YouTube videos isn't currently available from the dashboard's add area.

If your best content is in a video today, the practical workaround is to train on its substance directly: paste the key points, steps, or transcript excerpts into Train & Test Mode, or publish the transcript as a page and add its URL.

Text sources

Text sources are plain text content stored directly in your knowledge base, without a file or URL behind them. They appear under the Text filter tab in your source list. You'll typically see them created by integrations — for example, Help Scout Assist stores imported conversation threads as Text sources.

Closely related is the Q&A source type: every fact, policy, or instruction you teach the bot by chatting in Train & Test Mode is saved as a Q&A source. For ad-hoc knowledge — "our support hours are 9–5 CET," "we don't ship to PO boxes" — teaching the bot in chat is faster than preparing a file, and the saved facts are individually visible and deletable under the Q&A tab of your source list.

Common questions

Is there a file size limit?

Large PDFs can take longer to process and very large ones may fail — if a file errors, try splitting it into smaller documents. (The 1MB limit you may see mentioned elsewhere applies to CSV/text files of URLs, not to PDF uploads.)

Can I upload Word documents or spreadsheets?

Not directly — the uploader accepts PDF and plain text. Export Word documents to PDF first. For spreadsheet-style reference data, a page on your site or a plain-text summary usually trains better than a raw table.

How do I update a file the bot learned from?

Upload the new version, wait for it to show Processed, then delete the old source from the list. See Training Sources Overview for how removal works.

Can I paste a block of text as a source?

Use Train & Test Mode — tell the bot the information in chat and it's saved to the knowledge base immediately. For longer content, save it as a .txt file and upload it.

Was this article helpful?