Manage Agent Memory

In Agent Runtime, there are two key concepts related to how agents store and process information: knowledge and memory. While they may seem similar, they serve distinct purposes that are crucial for effective agent operation.

Knowledge vs Memory: A Human Analogy

Think about how humans process information:

Knowledge is like what you learned in school - it's the foundational information that stays relatively constant and can be referenced when needed. For example, knowing that Paris is the capital of France, or understanding how to solve algebraic equations. This information is stable and doesn't change frequently.

Memory is more like your personal experiences and recent interactions - it's dynamic and contextual. For example, remembering what you had for breakfast this morning, or recalling a conversation you had with a friend yesterday. This information is temporary and constantly updating.

How This Applies to Agents

In Agent Runtime:

• Knowledge is the pre-loaded information your agent has access to, like documentation, guidelines, or specific domain expertise you've provided. It's the stable foundation that helps the agent make informed decisions.
• Memory is the context from recent interactions and experiences. It helps the agent maintain coherent conversations and remember important details from ongoing interactions.

Just as a human uses both their learned knowledge and recent memories to have meaningful conversations, an agent needs both types of information processing to function effectively.