​
Knowledge Bases

Knowledge bases provide agents with specialized information beyond their training data. They help agents answer specific questions, follow guidelines, or work within defined domains.

​
What is a Knowledge Base?

Definition

A knowledge base is a structured collection of information that an agent can access to answer questions or make decisions.
Think of a knowledge base as a specialized reference library for your agent.

​
Why Use Knowledge Bases?

Domain Expertise

Provide specialized information for specific fields

Current Information

Include data beyond the agent’s training cutoff

Custom Guidelines

Define rules and policies for the agent to follow

Consistent Responses

Ensure agents provide standardized answers

​
Types of Knowledge Bases

​
1. Document Collections

Text documents, articles, guides, or manuals that agents can reference.

​
2. Structured Data

Databases, tables, or other structured formats that organize information systematically.

Examples

  • Product catalogs with specifications
  • Customer information databases
  • Statistical data collections
  • Reference tables for compliance information

​
3. Vector Databases

Special databases that store information as numerical representations (vectors) for semantic search.

​
Implementing Knowledge Bases in PraisonAI

Here’s an example of how to use a knowledge base with an agent: 
from praisonaiagents import Agent

agent = Agent(
    name="Knowledge Agent",
    instructions="You answer questions based on the provided knowledge.",
    knowledge=["small.pdf"]
)

agent.start("What is KAG in one line?")

​
Advanced Configuration

For more control over the knowledge base, you can specify a configuration: 
from praisonaiagents import Agent

config = {
    "vector_store": {
        "provider": "chroma",
        "config": {
            "collection_name": "custom_knowledge",
            "path": ".praison",
        }
    }
}

agent = Agent(
    name="Knowledge Agent",
    instructions="You answer questions based on the provided knowledge.",
    knowledge=["small.pdf"],
    knowledge_config=config
)

agent.start("What is KAG in one line?")

​
Knowledge Retrieval Process

When an agent uses a knowledge base, this typical process occurs:
  1. Query Processing: The user’s question is analyzed
  2. Search: The system searches the knowledge base for relevant information
  3. Ranking: Results are ranked by relevance
  4. Synthesis: The agent creates an answer using the retrieved information

​
Best Practices for Knowledge Bases

Keep Information Current

Regularly update your knowledge base

Organize Logically

Structure information in intuitive categories

Prioritize Quality

Focus on accurate, high-quality information

Include Examples

Add examples to illustrate complex concepts

​
When to Use Knowledge Bases

Knowledge bases are particularly valuable when:
  1. Your agent needs to reference specific information that may change over time
  2. You need to ensure consistent answers to common questions
  3. Your agent needs to follow specific guidelines or protocols
  4. You want to provide expertise in specialized domains
Start with a small knowledge base focusing on the most important information, then expand as needed.

​
Creating a Simple Knowledge Base

For beginners, you can start with a simple text-based knowledge base: 
# Company FAQ Knowledge Base

## Return Policy
Our return policy allows customers to return products within 30 days of purchase for a full refund.

## Shipping Information
Standard shipping takes 3-5 business days. Express shipping takes 1-2 business days.

## Product Warranty
All products come with a 1-year limited warranty covering manufacturing defects.
Save this as a text file and add it to your knowledge base: 
kb = KnowledgeBase()
kb.add_document("company_faq.txt")
In the next lesson, we’ll explore how agents handle tasks and the task management process.