📚 Knowledge Base Setup

Train your AI assistant with your content to provide accurate, contextual answers.

Overview

The Knowledge Base is the foundation of your AI assistant. It contains the content that the AI uses to answer customer questions. Better content means better answers.

Content Sources

🌐
Website Crawl

Auto-import from your docs, help center, or website

📝
Manual Articles

Create FAQs and articles directly in the dashboard

🔌
API Upload

Sync content programmatically via REST API

Website Crawling

The fastest way to populate your knowledge base. Point the crawler at your documentation site and it will automatically extract and index all content.

How to Crawl

  1. Go to Knowledge Base in your dashboard
  2. Click Crawl Website
  3. Enter your documentation URL (e.g., https://docs.yoursite.com)
  4. Configure crawl settings (depth, include/exclude patterns)
  5. Click Start Crawl
// Crawl via API
const response = await fetch('https://api.aetherai.support/api/kb/crawl', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    url: 'https://docs.yoursite.com',
    maxDepth: 3,
    includePatterns: ['/docs/*', '/help/*'],
    excludePatterns: ['/blog/*', '/changelog/*']
  })
});

const { jobId, status } = await response.json();

Crawl Settings

SettingDescription
Max DepthHow many links deep to follow (1-5)
Include PatternsOnly crawl URLs matching these patterns
Exclude PatternsSkip URLs matching these patterns
Respect robots.txtFollow robots.txt directives (recommended)

Manual Articles

Create FAQ articles directly in the dashboard. Perfect for common questions that aren't covered in your existing documentation.

Creating an Article

  1. Go to Knowledge BaseArticles
  2. Click New Article
  3. Enter a clear, descriptive title
  4. Write the content (Markdown supported)
  5. Add tags for categorization
  6. Click Publish

Tip: Write articles in a Q&A format. Use the question as the title and the answer as the content. This helps the AI match user questions more accurately.

API Upload

Sync content programmatically from your CMS, internal docs, or any other source.

// Create or update an article
const response = await fetch('https://api.aetherai.support/api/kb/articles', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    externalId: 'article-123', // Your unique ID for upsert
    title: 'How do I reset my password?',
    content: 'To reset your password, go to Settings > Security...',
    category: 'Account',
    tags: ['password', 'security', 'account'],
    metadata: {
      source: 'zendesk',
      lastUpdated: '2024-01-15'
    }
  })
});

const { id, status } = await response.json();
// Bulk upload articles
const response = await fetch('https://api.aetherai.support/api/kb/articles/bulk', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer YOUR_API_KEY',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    articles: [
      { title: 'Article 1', content: '...' },
      { title: 'Article 2', content: '...' },
      // Up to 100 articles per request
    ]
  })
});

AI Training

When you add content, the AI automatically processes and indexes it. Here's how the training works:

1
Content Extraction

HTML is parsed and cleaned. Boilerplate (navigation, footers) is removed.

2
Chunking

Content is split into semantic chunks for better retrieval accuracy.

3
Embedding

Each chunk is converted to a vector embedding for semantic search.

4
Indexing

Vectors are indexed for fast similarity search across your entire knowledge base.

Best Practices

✅ Do

  • Use clear, descriptive titles that match how users ask questions
  • Keep articles focused on a single topic
  • Include common variations of questions in the content
  • Update content regularly to keep it accurate
  • Use categories and tags for organization

❌ Avoid

  • Vague titles like "FAQ" or "Help"
  • Very long articles covering multiple topics
  • Duplicate content across articles
  • Outdated or incorrect information
  • Internal jargon that customers won't understand