Skip to main content

Uploading Documents

Upload PDF files, Word documents, CSV files, or text files to your Knowledge Base. Your agent will learn from these documents and use them to answer customer questions.

What File Types Work

Documents:
  • PDF files (.pdf)
  • Word documents (.docx)
  • Text files (.txt)
  • Markdown files (.md)
  • CSV files (.csv)
Best for:
  • Policies and guidelines
  • Product catalogs
  • FAQ documents
  • Support guides
  • Price lists
  • Specifications

Upload Steps

1. Go to Knowledge Base
  • Click your agent name
  • Click “Knowledge Base” in the sidebar
2. Click “Upload Document”
  • Button is at the top
  • Or drag and drop files directly
3. Choose Your File
  • Select from your computer
  • Multiple files at once (optional)
4. Give it a Name (optional)
  • Example: “Return Policy 2024”
  • Or use the default filename
5. Click Upload
  • Agent processes the document (1-2 minutes for typical files)
  • Status shows “Uploading” then “Ready”
6. Done!
  • Document is now in your Knowledge Base
  • Agent can immediately use it to answer questions

Tips for Success

Clear Names
  • Use descriptive names: “FAQ - Shipping 2024”
  • Not: “doc_final_v3_REAL.pdf”
Well-Organized Documents
  • Use headers and sections
  • Helps agent find answers faster
  • Example: ”### Return Policy”, ”### Shipping Info”
Keep It Current
  • Update prices when they change
  • Remove outdated policies
  • Delete old versions
Start Small
  • Upload 1-2 documents first
  • Test agent answers in Chatlab
  • Add more as needed

File Size Limits

  • Single file: Up to 50MB
  • Total storage: Depends on your plan
  • Most documents under 5MB anyway

What Happens Next

After upload:
  1. Processing - Agent reads and understands document (usually instant)
  2. Indexed - Document becomes searchable by agent
  3. Ready to use - Agent can answer questions using this content
You’ll see the document in your Knowledge Base list. Click it to see preview and stats.

Troubleshooting

Upload failed
  • Check file isn’t corrupted
  • Try different file format (PDF instead of Word)
  • Ensure file isn’t password protected
Agent not finding information
  • Check document has clear headings
  • Information might be embedded in images (upload text instead)
  • Use Train Your Agent to test

Next: Add Websites & URLs to crawl your website automatically Document Type/Category: Classify your document for better organization
  • Support/FAQ
  • Product Documentation
  • Company Policies
  • Technical Guides
  • Marketing Materials
Chunk Settings: Control how the document is split for searching (more details in the Document Chunking guide)
  • Default (Recommended): 300-500 tokens per chunk
  • Small Chunks: 200-300 tokens (for detailed Q&A style content)
  • Large Chunks: 500-800 tokens (for narrative content that needs more context)
Metadata: Add optional tags or labels
  • Tags help with organization and filtering
  • Example tags: “2024”, “premium-tier”, “technical”, “customer-facing”

Step 5: Upload and Process

  1. Review your settings
  2. Click “Upload” or “Start Processing”
  3. Watch the progress bar as your file uploads
  4. Processing begins automatically after upload completes

Step 6: Verify Upload Success

After processing completes (usually 30 seconds to 5 minutes depending on file size):
  1. Check the status indicator: Look for a green checkmark or “Active” status
  2. Preview the content: Click on the document to see how it was processed
  3. View chunks: See how the document was split into searchable pieces
  4. Test in Chatlab: Ask questions related to the document to verify it’s working

File Size Limits and Batch Upload

Single File Upload

  • Maximum file size: 10MB per file (for most accounts)
  • Larger files: Contact support for enterprise limits up to 50MB
  • Processing time: Approximately 1 minute per MB

Batch Upload

Upload multiple documents simultaneously to save time:
  1. Select “Batch Upload” option
  2. Choose multiple files (up to 20 at once)
  3. Configure shared settings (applied to all files)
  4. Upload and process all files together
Batch upload example: Upload your entire product manual series (10 PDFs) in one operation rather than individually.

Storage Limits

  • Starter Plan: 100MB total storage
  • Professional Plan: 500MB total storage
  • Enterprise Plan: Custom storage limits
  • Current usage: Check your dashboard to see available space

Document Processing Explained

Understanding what happens during processing helps you troubleshoot and optimize:

The Processing Pipeline

  1. Text Extraction: ChatCrafterAI extracts readable text from your document
    • For PDFs: OCR is applied if needed for scanned documents
    • For Word docs: Formatting is preserved where relevant
    • For CSV: Data is structured appropriately
  2. Structure Analysis: The system identifies headers, sections, and organization
    • Headings become important markers for chunking
    • Tables and lists are preserved
    • Page breaks and logical sections are noted
  3. Chunking: The document is split into searchable pieces (see Chunking guide)
    • Respects document structure
    • Maintains context across chunks
    • Creates overlaps to prevent information loss
  4. Embedding Generation: Each chunk is converted to a mathematical vector representation (see Embeddings guide)
    • Enables semantic search
    • Captures meaning beyond keywords
    • Powers the intelligent search capabilities
  5. Indexing: Processed chunks are added to the searchable database
    • Immediately available for agent queries
    • Integrated with existing Knowledge Base content
    • Ready for customer questions

Processing Time Expectations

  • 1-page document: 10-30 seconds
  • 10-page document: 1-2 minutes
  • 50-page document: 3-5 minutes
  • Large manual (200+ pages): 10-20 minutes
Complex documents with lots of images, tables, or special formatting may take longer.

Best Practices for Document Upload

1. Organize by Topic

Instead of uploading everything into one massive Knowledge Base, create separate, focused collections: Good approach:
  • Product Documentation KB (for technical users)
  • Customer Support KB (for general questions)
  • Internal Policies KB (for employee-facing agent)
Why this works: Helps the agent focus on relevant content and improves search accuracy.

2. Use Clear, Descriptive Headings

The AI uses your document structure to understand content: Good document structure:
# Return Policy

## Eligibility Requirements
- Items must be unused
- Original packaging required
- Within 30 days of purchase

## How to Initiate a Return
1. Log into your account
2. Navigate to Order History
3. Select the item to return
4. Follow the return wizard
Why this works: Clear headings help the chunking algorithm create meaningful, self-contained pieces.

3. Include Metadata and Context

Add dates, versions, and categories to your documents: Example:
  • Document name: “Shipping Policy - Updated January 2024”
  • Tags: “2024”, “policies”, “shipping”, “customer-facing”
  • Description: “Current shipping times, costs, and international options”
Why this works: Makes it easier to find and update documents later. When your shipping policy changes in June 2024, you know exactly which document to update or replace.

4. Remove Duplicates Before Uploading

Check if content already exists in your Knowledge Base:
  • Duplicate content can confuse the search system
  • The agent might return multiple similar chunks
  • Wastes storage space
Before uploading: Search your existing KB for similar content. If found, consider updating the existing document rather than adding a new one.

5. Clean Up Document Formatting

Before uploading, ensure documents are clean:
  • Remove unnecessary headers/footers from every page
  • Delete navigation elements (if from a web export)
  • Remove “Page X of Y” markers
  • Clean up any OCR errors in scanned documents
Why this works: Cleaner documents = cleaner chunks = more accurate search results.

Document Organization Examples

Example 1: Support FAQ Document Structure

# Frequently Asked Questions

## Account Management
### How do I reset my password?
[Detailed answer]

### How do I update my email address?
[Detailed answer]

## Billing Questions
### What payment methods do you accept?
[Detailed answer]

### How do I update my credit card?
[Detailed answer]

## Product Questions
### What's included in the Premium plan?
[Detailed answer]
This structure allows the agent to:
  • Find specific Q&A pairs quickly
  • Understand the category of each question
  • Provide complete, focused answers

Example 2: Product Manual Structure

# Widget Pro 3000 User Manual

## Getting Started
### Unboxing and Setup
### First-Time Configuration
### Basic Operation

## Advanced Features
### Feature A: Turbo Mode
### Feature B: Smart Scheduling
### Feature C: Remote Control

## Troubleshooting
### Device Won't Power On
### Connection Issues
### Error Codes
This structure enables:
  • Progressive disclosure (basics to advanced)
  • Easy troubleshooting lookups
  • Feature-specific queries

Troubleshooting Common Upload Issues

Issue: File Too Large

Symptoms: Upload fails with “File size exceeds limit” error Solutions:
  1. Compress PDF files (many tools available online)
  2. Split large documents into smaller sections
  3. Remove high-resolution images if not needed
  4. Upgrade to a plan with larger file limits

Issue: Format Not Supported

Symptoms: “Unsupported file format” error Solutions:
  1. Convert to PDF (universally supported)
  2. Export from your original application to a supported format
  3. Copy content to a Word or text document
  4. Check that file extension matches actual file type

Issue: Processing Failed

Symptoms: Upload completes but processing shows error Solutions:
  1. Check if PDF is password-protected (remove password first)
  2. Verify document isn’t corrupted (try opening it on your computer)
  3. Look for special characters in filename (use simple names)
  4. Try re-uploading the file
  5. Contact support if issue persists with specific file

Issue: Extracted Text Looks Wrong

Symptoms: Preview shows garbled or missing text Solutions:
  1. For scanned PDFs: Use better OCR software before uploading
  2. Export document to text format first, then upload text file
  3. Manually copy-paste content into a new document
  4. Check if document uses special fonts or encoding

After Upload: Next Steps

Once your documents are successfully uploaded:
  1. Test in Chatlab: Ask questions that should be answered by your uploaded content
  2. Review chunks: Check how the document was split (in document details)
  3. Refine if needed: Re-upload with better structure if results aren’t optimal
  4. Add related documents: Upload complementary content for comprehensive coverage
  5. Set up updates: Create a schedule for updating time-sensitive documents
Your uploaded documents form the foundation of your agent’s knowledge. Taking time to organize and structure them properly pays off in better, more accurate answers for your customers.