Step 5 of 1050% complete
4 min read

Training Your Agents (Knowledge Base)

Upload documents and scrape websites to train your AI agents with custom knowledge

Training Your Agents (Knowledge Base)

Train your agents with custom knowledge by uploading documents or scraping websites. Your agent can then answer questions based on this content.

How Training Works

When you upload training data:

  1. Content is chunked into smaller sections using AI
  2. Embeddings are created (semantic understanding)
  3. Agent gets a search_knowledge_base tool automatically
  4. When users ask questions, relevant chunks are retrieved
  5. Agent uses this context to answer accurately

Uploading Documents

Supported File Types

  • PDF - Reports, manuals, guides
  • DOCX - Microsoft Word documents
  • TXT - Plain text files

How to Upload

  1. Go to workspace → agent page
  2. Click Training tab
  3. Click "Upload Document" or drag & drop file
  4. Wait for processing (can take 30-60 seconds)
  5. Document appears in list when ready

File Limits

  • Max size: 5MB per file
  • Max pages: 20 pages (for PDFs)
  • Max chars: 50,000 characters
  • No limit on number of files (pay-as-you-go storage)

Best Practices

  • Upload well-organized files with clear headings
  • One topic per document - keeps responses accurate
  • Keep content fresh and updated regularly
  • Use descriptive file names

Web Scraping

Scrape content directly from websites to train your agent.

How to Scrape

  1. Go to Training tab
  2. Click "Scrape Website"
  3. Enter URL of the page
  4. Click "Scrape"
  5. Content is extracted and processed automatically

What Gets Scraped

  • Main content text
  • Headings and structure
  • Lists and bullet points
  • Excludes: Ads, navigation, footers, scripts

Web Scraping Costs

  • Base cost: $0.001 per page
  • Platform markup: Applied (typically 900%)
  • Your cost: ~$0.01 per page
  • You can add your own markup for clients

Example: Scraping 10 pages = ~$0.10

Use Cases

  • Company "About Us" and FAQ pages
  • Product documentation
  • Blog posts and articles
  • Help center content
  • Policy and legal pages

Managing Training Data

Viewing Documents

Training tab shows all uploaded documents:

  • File name and type
  • Upload date
  • Processing status (Processing, Ready, Failed)
  • Storage size

Deleting Documents

  1. Find document in list
  2. Click three-dot menu (⋮)
  3. Click "Delete"
  4. Confirm deletion

Deleting frees up storage space immediately.

Re-uploading

If you need to update content:

  1. Delete the old document
  2. Upload the new version

The agent will use the latest version automatically.

Storage & Costs

How Storage is Calculated

  • Storage is measured in MB (megabytes)
  • Charged based on actual usage
  • Billed daily at midnight
  • Base cost: $5/GB per month (~$0.16/day)

Example calculation:

  • 200MB of documents uploaded
  • 0.195 GB × $5/GB = $0.98/month
  • Platform markup (20%): $1.18/month to you
  • Your markup (30%): $1.53/month to client
  • Your profit: $0.35/month

Viewing Storage Usage

Check storage in:

  • Agency Settings → Credits & Billing → Storage Usage (total across all workspaces)
  • Workspace → Training tab → Storage summary (per workspace)

How Agents Use Training Data

Automatic Tool Injection

When you upload training data, agents automatically get:

  • search_knowledge_base tool
  • Ability to search uploaded documents
  • Semantic + keyword hybrid search

When to Use Training

Best for: Company-specific info, product docs, internal policies, FAQs, industry knowledge

Not needed for: General knowledge, current events, simple FAQs (use system instructions instead)

Troubleshooting

Upload fails?

  • Check file size (<5MB)
  • Verify file type (PDF, DOCX, TXT only)
  • Ensure file isn't password-protected or corrupted

Agent not using training data?

  • Wait for "Ready" status on documents
  • Test with specific questions about uploaded content
  • Check system instructions don't override knowledge base

Processing takes too long?

  • Large files (>2MB) take 30-60 seconds
  • PDFs with images process slower
  • Try splitting into smaller files

Next Steps

Your agent is now trained with custom knowledge:

  1. Add Tools & Integrations - Connect external services
  2. Test in Playground - Verify training worked
  3. Deploy Your Agent - Share with users

Ready to Build AI Agents?

Join agencies across the UK using our white-label platform to deliver AI solutions to their clients.