Training Your Agent

Process and index your knowledge sources so your AI agent can answer questions accurately

Training Your Agent#

After adding knowledge sources (files, websites, Q&A pairs, and more), you need to train your agent to process and index the content. Training converts raw content into searchable, AI-ready knowledge.

What Happens During Training#

Training involves multiple automated steps that run in the background.

1. Content Extraction#

Content is extracted from each source type:

• Files -- Text from PDFs, Word docs, Excel spreadsheets, and more
• Websites -- Downloaded and cleaned HTML pages
• Google Drive -- Synced and extracted from linked documents
• Facebook/Instagram -- Parsed imported message history
• Q&A -- Formatted question-answer pairs
• Structured Data -- Processed records from Google Sheets

2. Intelligent Chunking#

Content is intelligently split into small, searchable pieces. The system respects natural boundaries like paragraphs and sentences to preserve context and meaning.

3. Content Organization#

Chunks are organized into categories based on content type so the AI can prioritize the right information for each question:

Each category has a configurable priority that determines how likely its content is to be retrieved. The system automatically prioritizes the right category based on the question type.

4. Embedding Generation#

Each chunk is converted into a form the AI can search semantically. This means "How much?" will match content about "price" even though the exact words are different.

5. Indexing#

Processed content is stored and indexed with metadata (source, position, priority, recency) so it can be retrieved quickly during conversations.

How to Train Your Agent#

Step 1: Add Knowledge Sources#

Before training, add at least one source:

• Upload files (PDF, Word, etc.)
• Add a website to crawl
• Sync Google Drive documents
• Import Facebook or Instagram data
• Create Q&A pairs
• Paste text directly
• Add structured data from Google Sheets
• Create time-sensitive promos or events

Tip: Add multiple sources at once before training. Batch processing is more efficient than training one source at a time.

Step 2: Click "Train Agent"#

1. Go to your agent dashboard
2. Click Sources in the sidebar
3. Click the Train Agent button (top-right)
4. Confirm to start training

Step 3: Monitor Progress#

Training runs in the background:

• Real-time progress bar showing chunks processed
• Step indicators showing the current stage (extraction, chunking, embedding)
• Time estimate showing approximate remaining time

Typical processing times:

• 10 pages: approximately 1-2 minutes
• 100 pages: approximately 5-10 minutes
• 1,000 pages: approximately 30-60 minutes

Step 4: Training Complete#

When finished, you will see:

• "Training Complete" status
• Total chunks created (number of searchable pieces)
• Sources trained (count of successfully processed sources)

Your agent is now ready to answer questions using the trained content.

When to Re-Train#

Re-train your agent when:

Adding New Sources#

• Uploaded new files
• Added website pages
• Imported updated Facebook or Instagram data
• Created new Q&A pairs

Editing Existing Sources#

• Modified Q&A answers
• Updated text sources
• Re-crawled a website (new content)

Deleting Sources#

• Removed outdated files
• Deleted old Q&A pairs
• Archived unused sources

Changing Configuration#

• Adjusted retrieval settings
• Changed priority scores

Note: Training is incremental. Only new or changed sources are reprocessed, so re-training is faster after the initial run.

Understanding Training Status#

Sources have different status indicators:

Training Performance#

Optimization Tips#

1. Batch Training

• Add multiple sources at once, then train once
• Much faster than training sources individually

2. Use Smaller Files

• Split very large PDFs into smaller sections
• Faster processing and easier to update specific sections later

3. Pre-Clean Content

• Remove unnecessary pages (covers, table of contents, indexes)
• Delete duplicate content across sources
• Fix formatting issues before upload

4. Schedule Large Imports

• Run large imports (1,000+ pages) during low-traffic periods
• Avoid peak hours for faster processing

Troubleshooting Training Issues#

"Training failed"#

Common causes:

1. File corrupted -- Re-upload the file
2. Website unreachable -- Check the URL and try again
3. Timeout -- File too large, split into smaller files

Solution: Click "Retry Training" on the failed source.

"Training stuck"#

Possible issues:

• Large file still processing (be patient)
• Worker overload (try again later)
• Network issue (check your connection)

Fix: Wait 10 minutes, refresh the page, and check the status.

"Content not appearing in chat"#

Checklist:

• Training completed successfully (status shows "Trained")
• Content matches query (test with exact phrases from the source)
• Source is enabled and not archived

Debug: Test in the playground with exact text from your source to verify retrieval.

"Too many chunks created"#

Symptoms:

• Source created a very large number of chunks
• Irrelevant content being returned

Solutions:

1. Remove duplicate content from the source
2. Split the source into smaller, focused files
3. Use URL pattern filtering for website sources

"Training uses too many credits"#

Reduce costs:

1. Remove duplicate sources
2. Pre-process files to remove unnecessary content
3. Split large files into focused sections covering specific topics

Training Best Practices#

1. Quality Over Quantity#

• 10 high-quality sources outperform 100 low-quality sources
• Focus on relevant, accurate, up-to-date content
• Remove outdated information regularly

2. Organize by Topic#

• Group related sources (e.g., "Pricing", "Shipping", "Returns")
• Use consistent naming conventions
• Makes debugging retrieval issues much easier

3. Test After Training#

• Ask questions to verify the AI retrieves the right content
• Check that answers are accurate and complete
• Adjust sources if anything is missing or incorrect

4. Keep Sources Updated#

• Re-train monthly for static content
• Re-train weekly for frequently changing content
• Set reminders for website re-crawls

5. Monitor Training Logs#

• Check for failed sources after each training run
• Review warning messages
• Fix issues promptly before they affect conversations

Next Steps#

• Testing Your Agent -- Verify training results with thorough testing
• Web Widget -- Add your agent to your website
• Understanding Agents -- Learn how retrieval and AI responses work