AI Assistants

Create deployable AI assistants trained on your content. Assistants are perfect for customer-facing chatbots, knowledge bases, and interactive support systems.

What are Assistants?

Assistants are AI chatbots that you can train on your own content and deploy across messaging platforms. They differ from Agents in their focus:

Creating an Assistant

Step 1: Access the Assistant Builder

1. Go to Dashboard > Assistants
2. Click Create New Assistant

Step 2: Basic Configuration

• Name: Give your assistant a friendly name (e.g., "Support Bot", "Product Guide")
• Welcome Message: The first message users see
• Avatar: Choose a template or upload custom image
• Color Theme: Select brand colors for the chat widget

Step 3: Choose a Template

Select a visual template for how your assistant appears:

Step 4: Train on Your Content

Upload content to train your assistant:

Supported Content Types

• Documents: PDF, Word, text files
• Websites: URLs to scrape and learn from
• Q&A Pairs: Direct question-answer mappings
• Knowledge Base: Structured help articles

Training Methods

Method 1: File Upload 1. Click Add Training Data 2. Upload documents (PDF, DOCX, TXT) 3. Wait for processing 4. Assistant learns from content

Method 2: Website Import 1. Click Import from URL 2. Enter website URL 3. Select pages to include 4. Content is extracted and learned

Method 3: Manual Q&A 1. Click Add Q&A Pair 2. Enter common questions 3. Provide ideal answers 4. Build a FAQ knowledge base

Step 5: Configure Behavior

Personality Settings: - Tone: Professional, Friendly, or Casual - Response Length: Concise, Balanced, or Detailed - Creativity: Focused (factual) to Creative (imaginative)

Fallback Behavior: - What to say when the assistant doesn't know the answer - Option to collect user email for follow-up - Handoff to human support

Step 6: Test Your Assistant

Use the Preview panel to test conversations:

1. Send sample questions
2. Verify responses are accurate
3. Check edge cases
4. Refine training data as needed

Step 7: Deploy

Once satisfied, deploy to your channels:

• Chat Widget: Embed on your website
• Telegram: Connect as a bot
• Discord: Add to your server
• Slack: Install to workspace
• API: Integrate programmatically

Assistant Statuses

Managing Training Data

Adding More Content

1. Go to your assistant's settings
2. Click Training Data tab
3. Add new files, URLs, or Q&A pairs
4. Click Retrain to update the model

Removing Content

1. Go to Training Data tab
2. Find the content to remove
3. Click the delete icon
4. Retrain to apply changes

Best Practices for Training

1. Quality over quantity - Clean, accurate content trains better
2. Cover edge cases - Include variations of common questions
3. Update regularly - Keep information current
4. Test after training - Verify responses are still accurate

Chat Widget Integration

Embed your assistant on any website:

Quick Embed

Copy the embed code from your assistant's deploy settings:

<script src="https://neurogen.cc/widget/assistant/{ASSISTANT_ID}.js"></script>

Customization Options

<script>
  NeurogenWidget.init({
    assistantId: 'YOUR_ASSISTANT_ID',
    position: 'bottom-right', // or 'bottom-left'
    primaryColor: '#22d3ee',
    welcomeMessage: 'Hi! How can I help?',
    placeholder: 'Type your question...'
  });
</script>

Analytics & Insights

Track your assistant's performance:

Key Metrics

• Total Conversations: Number of chat sessions
• Messages: Total messages exchanged
• Satisfaction Rate: User feedback scores
• Resolution Rate: Questions answered without escalation
• Popular Topics: Most asked questions
• Credit Usage: Token consumption

Using Analytics

1. Go to your assistant's Analytics tab
2. Select date range
3. Review metrics and trends
4. Export data for further analysis

Improving Performance

Based on analytics: - Add training data for unanswered questions - Refine responses for low-satisfaction topics - Update outdated information - Optimize for popular questions

Advanced Features

Conversation Variables

Capture and use user information:

Welcome back, {user_name}! How can I help you today?

Conditional Responses

Set up responses based on context: - Time of day greetings - User segment targeting - A/B testing responses

Integrations

Connect to external systems: - CRM (capture leads) - Help desk (create tickets) - Calendar (book appointments) - Custom webhooks

Troubleshooting

Assistant Gives Wrong Answers

1. Review and update training data
2. Add specific Q&A pairs for problem areas
3. Check for conflicting information in sources
4. Retrain the assistant

Slow Response Times

1. Check your credit balance
2. Reduce response length setting
3. Optimize training data size
4. Contact support if persistent

Widget Not Appearing

1. Verify the embed code is correct
2. Check for JavaScript errors in browser console
3. Ensure assistant is set to Active
4. Try clearing browser cache

Low User Engagement

1. Improve welcome message
2. Add proactive conversation starters
3. Optimize widget position
4. A/B test different approaches

Pricing & Credits

Assistants use credits based on: - Input tokens: User messages - Output tokens: Assistant responses - Training: Processing uploaded content

Monitor usage in the Analytics tab and Billing section.

Assistants vs. Agents: When to Use Which

Choose Assistants when: - Building customer-facing FAQ bots - Need to train on specific content - Want quick setup with templates - Embedding on websites

Choose Agents when: - Need multi-step task automation - Integrating with tools and APIs - Building complex workflows - Requiring programmatic control

You can use both together - Assistants for front-line support, Agents for complex escalations.

Next Steps

• Create an Agent - Build task-automation agents
• Telegram Integration - Deploy to Telegram
• Discord Integration - Deploy to Discord
• API Reference - Programmatic access

Need help? Contact support@neurogen.cc