Skip to content

Chatbot Builder

Create custom AI chatbots trained on your company's knowledge for customer support, sales assistance, and self-service.

Overview

The Chatbot Builder lets you create branded AI chatbots that understand your products, services, and company context. Deploy them on your website or integrate via embed codes.

Key Features

Custom Training Data

Train your chatbot on multiple data sources:

Source Type Description Formats
Documents Upload files for automatic extraction PDF, XLS, XLSX, CSV
Website Crawl Automatically extract content from URLs Single page or full site crawl
Q&A Pairs Define specific question-answer pairs Text input
Text Content Add raw text content directly Text input

Brand Customization

Configure your chatbot's appearance and behavior:

  • Custom Logo — Upload your company logo
  • Color Theme — Match your brand colors
  • First Message — Set the greeting message
  • Instructions — Define the chatbot's role and behavior
  • Position — Choose widget placement (left/right)

Multi-Domain Support

Deploy your chatbot across multiple domains:

  • Add unlimited domains per chatbot
  • Domain validation for security
  • Cross-domain embedding support

AWS-Native Knowledge Base

Training data is processed using AWS-native services:

  • Amazon Nova Embeddings — Vector embeddings generation
  • S3 Storage — Secure vector storage
  • FTR Compliant — All data stays within AWS

Getting Started

Step 1: Create a Chatbot

  1. Navigate to Settings > Chatbot Training
  2. Click Create New Chatbot
  3. Enter a title for your chatbot
  4. A new chatbot is created with default settings

Step 2: Configure Settings

  1. Click on your chatbot to open settings
  2. Set the Title and Role (e.g., "Customer Support Agent")
  3. Write Instructions that define behavior:
You are a helpful customer support agent for Acme Corp.
You assist customers with product questions, troubleshooting,
and account inquiries. Be friendly, professional, and concise.
If you don't know something, offer to connect them with a human agent.
  1. Set the First Message greeting:
Hi! I'm the Acme Support Assistant. How can I help you today?
  1. Choose an AI Model (Claude Sonnet 4, Nova Lite, etc. — available models depend on your plan)
  2. Upload a Logo (optional)

Step 3: Add Training Data

Upload Documents

  1. Click the Documents tab
  2. Click Upload File
  3. Select PDF, Excel, or CSV files
  4. Files are automatically parsed and content extracted
  5. Click Train to process the documents

Crawl Website

  1. Click the Website tab
  2. Enter your website URL (e.g., https://docs.acme.com)
  3. Choose Single Page or Full Site Crawl
  4. Click Crawl to extract content
  5. Review extracted pages
  6. Click Train to process

Add Q&A Pairs

  1. Click the Q&A tab
  2. Enter a Question that customers might ask
  3. Enter the Answer you want the chatbot to give
  4. Click Add
  5. Repeat for common questions
  6. Click Train to process

Add Text Content

  1. Click the Text tab
  2. Enter a Title for the content
  3. Enter the Text content
  4. Click Add
  5. Click Train to process

Step 4: Deploy

Get Embed Code

  1. Navigate to Chatbot Domains
  2. Add your domain (e.g., www.acme.com)
  3. Copy the embed code:
<script src="https://vell.ai/chatbot/embed.js"
        data-chatbot-id="YOUR_CHATBOT_ID"
        data-position="right">
</script>
  1. Add the script to your website's HTML

Configure Widget Settings

  1. Go to Settings > Floating Chat Settings
  2. Select your chatbot template
  3. Configure:
  4. Position — Left or right side of screen
  5. Login Required — Require users to log in
  6. Rate Limit — Messages per minute
  7. Show Timestamp — Display message times

Training Data Best Practices

Document Uploads

Optimize PDFs

  • Use text-based PDFs (not scanned images)
  • Ensure clear formatting
  • Remove unnecessary headers/footers

Excel/CSV Structure

  • First row should be headers
  • Keep data well-organized
  • Use clear column names

Website Crawling

Crawl Strategy

  • Start with documentation pages
  • Exclude login/account pages
  • Review extracted content for quality

Large Sites

For sites with 100+ pages, consider crawling specific sections rather than the entire site.

Q&A Pairs

Effective Q&A

  • Use actual customer questions from support tickets
  • Include variations of common questions
  • Keep answers concise and helpful

Text Content

Good Uses

  • Product specifications
  • Pricing information
  • Policy documents
  • FAQs

Training Status

After adding data, each item shows a status:

Status Description
Waiting Data added, not yet processed
Training Currently being processed
Trained Successfully added to knowledge base
Failed Processing error (retry or check content)

Chatbot Settings Reference

Setting Description Default
Title Internal name for the chatbot "Untitled chatbot"
Role Description of chatbot's purpose None
Model AI model used for responses Default model
First Message Greeting shown when chat opens None
Instructions System prompt defining behavior None
Logo Custom avatar image Default icon
Color Widget accent color Brand default
Width/Height Widget dimensions Auto

External Chat Settings

For embedding chatbots on external sites:

  1. Navigate to Settings > External Chat Settings
  2. Configure:
  3. Select the chatbot template to use
  4. Set rate limits for external users
  5. Enable/disable login requirements
  6. Add allowed domains in Chatbot Domains

Analytics

Track chatbot performance:

  • Total Conversations — Number of chat sessions
  • Messages — Total messages exchanged
  • Active Users — Unique users interacting
  • Popular Questions — Most common queries

Navigate to your chatbot's analytics to view metrics.

Troubleshooting

Chatbot Not Responding

  1. Check that training is complete (status: "Trained")
  2. Verify the chatbot is active
  3. Check your AI model API key configuration

Poor Response Quality

  1. Add more training data
  2. Improve instructions with clearer guidelines
  3. Add specific Q&A pairs for problem areas

Embed Not Loading

  1. Verify domain is added to allowed list
  2. Check for JavaScript errors in browser console
  3. Ensure HTTPS is used on both sites

Support

  • Email: support@vell.io
  • Documentation: FAQ
  • In-App Chat: Available in dashboard