Module 6: RAG & Knowledge Management

You’ve built your chatbots. Now let’s talk about how they work and how to manage your knowledge bases efficiently.

What is RAG?

RAG = Retrieval-Augmented Generation

When you ask a chatbot a question, it doesn’t “know” everything in its knowledge base. Instead:

1. Retrieval: Your question + KB files are processed together
2. Augmentation: Relevant information is pulled from KB
3. Generation: AI generates response using that context

User Question → [Retrieval] → Relevant KB Content → [Generation] → Response

Why This Matters

Token Estimation

Rough guidelines:

• 1 token ≈ 4 characters of English text
• 1 page of text ≈ 1,000-1,500 tokens
• Average email ≈ 100 tokens

Example: 10 markdown files × 2,000 tokens each = 20,000 tokens

Managing KB Size

Strategy 1: Chunking

Split large documents into focused sections:

# Instead of one large file:
scrapy-complete-guide.md (10,000 tokens)

# Split into focused files:
scrapy-selectors.md (1,500 tokens)
scrapy-spiders.md (2,000 tokens)
scrapy-settings.md (1,500 tokens)

Strategy 2: Selective Upload

Only upload relevant files per chatbot:

• scrapy-tutor → Scrapy docs only
• zapier-specialist → Zapier help content only

Strategy 3: Local Storage

Keep full KB locally, upload subsets:

my-chatbot-bundles/
├── scrapy-tutor/
│   ├── directive.md
│   └── knowledge-base/          # Full Scrapy docs (local)
│       ├── selectors.md
│       ├── spiders.md
│       └── ...
├── zapier-specialist/
│   ├── directive.md
│   └── knowledge-base/          # Full Zapier content (local)
│       ├── troubleshooting.md
│       ├── getting-started.md
│       └── ...

RAG Storage Limits

Practical tip: Keep knowledge bases under 5MB for free tier.

Versioning Your Knowledge

As content changes, maintain versions:

knowledge-base/
├── v1/                          # Initial scrape
│   ├── article-01.md
│   └── ...
├── v2/                          # Updated content
│   ├── article-01.md
│   └── ...
└── current/                     # Active version
    ├── article-01.md
    └── ...

When to Update KB

• Service introduces new features
• Major troubleshooting patterns emerge
• Your bot gives outdated information

Key Takeaways

1. RAG = Retrieval-Augmented Generation - How chatbots use KB
2. Token limits exist - Manage KB size carefully
3. Local storage + selective upload = efficient management
4. Version your content - Track changes over time

Next Steps

Important consideration: your data on cloud services. Continue to Module 7: Data Privacy Considerations