Add files, Text Snippets and Q&A

Library holds four kinds of items. Pick the one that matches the knowledge you have.

Upload a file#

1. Open Library and select Create Library Item. Type is already set to File.
2. Drop a file into the File box, or select it to browse.
3. Optionally add a Title, Tags and a Description of what the file contains and when to use it. For images, add Alt Text.
4. Select Upload File.

Each file can be up to 10 MB. Supported files include PDF, DOCX, PPTX, XLSX, plain text, Markdown, CSV and TSV, JSON, XML, YAML, HTML and EPUB, plus JPEG, PNG, GIF, WebP, HEIC/HEIF, AVIF, BMP and TIFF images. Images are converted when needed, so photos straight from a phone work.

A file's Title can be up to 200 characters, its Description up to 1,000 and its Alt Text up to 500.

Upload several files at once#

On the Library page, select Upload files, or drop or paste files onto the page. You can add up to 5 files at a time, and each becomes its own Library item. A file that's already in Library is skipped. Edit titles and details afterwards.

Write a Text Snippet#

1. In Create Library Item, set Type to Text Snippet.
2. Give the snippet a Title, choose its Format (Plain text, Markdown or HTML) and paste the Content.
3. To keep related snippets together, select Add another text snippet.
4. Select Create Text Snippet.

One item holds up to 500 snippets, and each snippet's title can be up to 200 characters.

Add Q&A entries#

1. In Create Library Item, set Type to Q&A.
2. Fill in the Question, any Alternate questions customers might ask instead (one per line), and the trusted Answer.
3. Select Add another Q&A for each extra entry.
4. Select Create Q&A.

One item holds up to 500 entries, and each entry up to 100 alternate questions.

Check what was extracted#

Every item is prepared for AI before agents can use it. Extraction starts on its own when you create or upload an item, and its status moves from Pending to Processing to Ready. From the row menu:

• View content shows the item and the text the agent reads from it.
• Edit changes the item. When you change a website's links or settings, or the text of a Text Snippet or Q&A, saving extracts it again on its own.
• Extract content runs the extraction again by hand.
• Extraction summary shows the extraction's status and any error.
• Delete removes the item.
• Files also have Download and Get Signed URL.

Get Signed URL creates a temporary link to a file. Set Disposition to Inline to open it in the browser, or Attachment to download it. The link works for 6 hours.

To find items, search the Library, or filter by type (Websites, Text Snippets, Q&A, Files), by extraction status (Pending, Processing, Ready, Failed), by Conversation (where a file was added: AI Chat, Agent test chat, Visitor chats) and by Deployment. A filtered list counts the items that match. See Fix and refresh Library items.

See where an item came from#

The last line under each item's name says where it came from:

• Added by a workspace member, for items added on the Library page.
• Added by a member in AI Chat, linking to the chat, or in an agent test chat, for files attached there.
• From a visitor on a deployment, for files a visitor sent in the chat. It links to the conversation in Inbox, and Visitor’s files lists everything that visitor sent.

Images an AI agent creates read like their conversation's files. A visitor's file stays with that visitor: the AI agent reads it in that visitor's own conversations, and other visitors and your team's chats never find it.

Label items with tags#

Use Tags to label items by topic or audience, such as returns or internal. Tags show on each Library row. An item can have up to 100 tags of up to 100 characters each.

Tags don't change what an agent can read. To keep internal items away from a customer-facing agent, leave them out of its Pick items selection. See Choose what each AI agent can read.