The Knowledge Base (KB) module enables you to structure, manage, and train the information used by your chatbot.
It provides a complete workflow to create content repositories, organize them into folders, upload Markdown files, tag content, and train a vector database for semantic retrieval (RAG).

This system ensures that your chatbot answers are grounded, consistent, and based exclusively on your curated data.

Overview #

Each chatbot can use one or more Knowledge Bases.

A Knowledge Base contains:

• Folders (logical grouping)
• Markdown files (content)
• Tags (metadata)
• Search & ordering tools
• Integration type configuration
• Vector database training

In the Chatbot Editor the developer can select:

1. The Knowledge Base
2. The specific folder

This allows precise context scoping and improves answer quality.

Create a Knowledge Base #

Steps #

1. Go to Chatbot Studio → Knowledge Base
2. Click Create Knowledge Base
3. Provide:
   • Name
   • Description (optional)
   • Integration Type
4. Save

Best Practices #

• Create one KB per domain or department
  Examples:
  • Sales
  • Technical Docs
  • Legal
  • HR
  • Product Manuals
• Avoid mixing unrelated content inside the same KB

Folder Management #

Folders help organize content logically and improve retrieval accuracy.

Create a Folder #

1. Open your Knowledge Base
2. Click New Folder
3. Name the folder
4. Save

Common Structures #

Product-based #

• Product A
• Product B
• FAQ

Department-based #

• Policies
• Procedures
• Training

Time-based #

• 2024
• 2025
• Archive

Recommendations #

• Keep folders focused
• Avoid overly large folders
• Split content by topic for better embedding precision

File Management #

Each folder contains Markdown (.md) files.

Markdown provides clean formatting and structured knowledge ingestion.

Supported Features #

• Headings (#, ##, ###)
• Lists
• Tables
• Links
• Code blocks
• Images (URLs or references)
• Plain text

Add a File #

1. Open folder
2. Click Upload or Create File
3. Paste or write Markdown content
4. Save

Example #

# Return Policy

## Time Limits
Customers can return items within 30 days.

## Conditions
- Product must be unused
- Original packaging required

Tips #

• Use clear headings
• Split long documents into multiple files
• Keep each file topic-specific (1 concept per file)
• Avoid mixing unrelated topics

Tagging System #

Tags provide metadata for filtering, search, and training refinement.

Add Tags #

Tags can be added:

• At file level
• At folder level
• At KB level

Examples #

• faq
• onboarding
• legal
• pricing
• internal
• public

Benefits #

Tags allow:

• Faster search
• Selective training
• Scoped retrieval
• Cleaner organization

Search #

The search tool helps quickly locate content inside a KB.

Capabilities #

• Full-text search
• Tag filtering
• Folder filtering
• Instant results

Use Cases #

• Find specific files
• Verify duplicates
• Audit knowledge
• Locate outdated content

Ordering & Priority #

You can define manual ordering of Files

How to Reorder #

• User the Order button in the top right menu

When to Use #

• Highlight critical documents
• Promote FAQs
• De-prioritize legacy content

Integration Types #

Each Knowledge Base can be connected to different ingestion strategies.

Available Types (examples) #

Manual #

Upload and manage Markdown manually.

File Sync #

Sync from:

• PDF

API #

Push content dynamically via API.

Web Scraping #

Import content from websites or documentation portals.

Selection Guidelines #

Scenario	Recommended
Static docs	Manual
Live documentation	Sync
Automated systems	API
Public site FAQ	Scraper

Vector Database Training #

Training converts your content into embeddings for semantic search.

Without training, the chatbot cannot use the content.

Train the KB #

1. Click Train Knowledge Base
2. System:
   • Splits text into chunks
   • Generates embeddings
   • Stores them in the vector database
3. Status updates shown in dashboard

When to Retrain #

• After adding files
• After editing content
• After deleting content
• After tag or structure changes

Best Practices #

• Train only finalized content
• Avoid frequent small retrains
• Batch updates for efficiency

Using Knowledge Bases in Chat #

When configuring or starting a chat:

Step 1 – Select Knowledge Base #

Choose the KB that contains the relevant information.

Step 2 – Select Folder (optional) #

Restrict retrieval to a specific folder.

Benefits #

• Improves precision
• Reduces hallucinations
• Speeds up retrieval
• Avoids cross-domain mixing

Example #

Task	KB	Folder
Sales question	Sales Docs	Pricing
Technical support	Tech Docs	Troubleshooting
HR policy	HR	Policies

Recommended Workflow #

1. Create KB
2. Create folders
3. Upload Markdown files
4. Add tags
5. Review structure
6. Train vector DB
7. Use in chat

Performance Tips #

• Keep files under 1–2k words each
• Use meaningful headings
• Avoid PDFs when possible (Markdown preferred)
• Remove duplicates
• Use tags consistently
• Train periodically

Common Issues #

Chatbot does not answer from KB #

→ Check if training was completed

Answers are inaccurate #

→ Split large files, add headings, retrain

Slow retrieval #

→ Reduce KB size or narrow folder scope

Too many irrelevant answers #

→ Use tags or folder filtering