Starting a Conversation

When a visitor opens the chat, they see:

1. Greeting message -- The first message from your assistant, configured in Settings.
2. Starter questions -- Clickable buttons below the greeting (up to 5). Click one to send it as a first message instantly.
3. Message input -- Type a question at the bottom and press Enter to send.

Use Shift+Enter to add a new line without sending.

Real-Time Streaming

AI responses stream in word by word. You can read the answer as it's being generated. The input is disabled while the assistant is responding.

Rich Formatting

AI responses support Markdown formatting automatically:

• Headers in different sizes
• Bold and italic text
• Bulleted and numbered lists
• Code snippets and code blocks
• Clickable links

Knowledge Citations

When the assistant uses content from your knowledge base, it shows a Sources section below the response. Each source includes:

• The content title
• A thumbnail image (if available)
• A link to the original source (web page, YouTube video, etc.)

Click a source to open it in a new tab. Sources only appear for content that has citations enabled. Content with citations turned off is still used to generate answers but won't show up in the Sources section.

You can control citations per content item in Building Knowledge.

File Upload

File upload is available for internal (knowledge base) and contract assistants. Public assistants do not support file upload.

How to Upload

1. Click the + button next to the message input.
2. Select one or more files.
3. Or drag and drop files directly onto the chat input area.

Supported File Types

Limits

• Maximum 10 MB per file.
• Up to 3 files at once.

How It Works

Selected files appear as compact chips below the input, showing the file name and an upload spinner. Files upload in the background while you type your message.

When you send, the file contents are included as context for the AI. If uploads are still in progress, the message waits until all uploads complete. A file attachment card shows the file type, name, and size in the conversation.

Smart Follow-Up Questions

When enabled on an assistant (in Settings), the AI generates contextual follow-up suggestions after each response.

• 2--3 clickable buttons appear below each AI response.
• These are questions the AI thinks you might want to ask next, based on the conversation so far.
• Smart questions only appear after the response is fully streamed.

Two types of smart questions:

• General -- Regular follow-ups that continue the current topic.
• Lead capture -- Call-to-action buttons linked to lead forms (public assistants only). For example, after a pricing question, the AI might suggest "Get a personalized quote."

Thread History

For Internal Assistants

A thread history panel is available on the left side of the chat:

• See your previous conversation threads.
• Click a thread to resume it.
• Each team member sees only their own threads.

For Public Assistants (Admin View)

Admins can view all visitor conversations from the Threads page in the sidebar. The threads table shows:

• Thread ID and title
• User avatar or country flag (for visitors)
• Message count and detected language
• Created and updated dates
• A message volume chart at the top

Click a thread to view the full conversation with all messages and timestamps.

Starting a New Thread

Click the New thread button (refresh icon) in the header to start a fresh conversation. Your previous thread is preserved in the history.

Preview Mode (Embedded Chat)

When the chat is embedded on a website in preview mode:

• New visitors see the greeting message, starter questions, and a "Start chat" button with an animated border.
• Returning visitors see the assistant icon and name, "Last message X ago" timestamp, and a "Continue chat" button alongside starter questions labeled "or ask something new."
• On mobile (screen width 700px or less), preview mode is the default for inline widgets. The chat opens fullscreen when the visitor taps to start.

Conversation Persistence

• Embedded chat (website visitors) -- Thread ID is stored in the browser's local storage. Returning visitors continue their conversation. Clearing browser data starts a fresh thread.
• Dashboard chat (team members) -- Threads are stored server-side and linked to your user account. They persist across devices and browsers.

50+ Languages, Automatically

The AI detects the visitor's language automatically and responds in the same language. Over 50 languages are supported. The language is detected from the first message and stays consistent throughout the conversation.

Keyboard Shortcuts

Error Handling

• Message limit reached -- If your organization has used all available messages for the billing period, the chat displays an error. Upgrade your plan to continue.
• Connection issues -- If the connection is interrupted during streaming, the partial response is preserved. The input becomes available again so you can retry.

Tips for Better Conversations

• Be specific. "What are the shipping options for orders over $100?" works better than "Tell me about shipping."
• Ask follow-ups. The AI remembers the full conversation context, so you can build on previous answers.
• Use starter questions. They are designed to showcase the assistant's best capabilities.
• Upload relevant files. For internal assistants, uploading a document lets the AI answer questions about it immediately.