UNDERSTANDING CONTEXT AND KNOWLEDGE BASE

What is Context?

Context is all the information the AI has at the moment it processes a request.

Context can come from:

• System Prompt;
• the conversation;
• Knowledge Base;
• files uploaded by the user;
• data from Tools;
• data of the current task.

Put simply:

Model = general capability. Context = what the AI currently knows about the problem at hand.

The same model with different context can produce very different results.

What is a Knowledge Base?

A Knowledge Base is a collection of knowledge prepared for an Agent to use repeatedly.

For example:

• methods;
• principles;
• processes;
• standards;
• domain expertise;
• glossary;
• internal documents;
• experience that has been systematized.

A KB is not a warehouse for all of your files.

It should only contain what the Agent truly needs to know to do its job.

A longer KB does not mean a better one

Providing more documents does not automatically make the Agent smarter.

A KB that is too large can cause:

• wrong retrieval;
• duplicated information;
• contradictions;
• larger context;
• higher cost;
• dilution of important principles.

The goal is:

enough knowledge as needed, not the most knowledge possible.