Skip to content

Understanding Memory in Our AI Platform

Note: The memory feature is still experimental and should be used with caution in production workloads.

What Is Memory for AI Agents?

At Wabee, Memory refers to the ability of AI Agents to store and recall information to enhance their interactions and performance. Memory allows agents to maintain context, learn from past experiences, and improve over time. There are two primary types of memory in our platform:

  1. Session Memory: Memory associated with a single interaction or conversation session.
  2. Long-Term Memory: Memory associated with the agent itself, persisting across multiple sessions and over its lifetime.

These two types of memory are interconnected, working together to provide a seamless and intelligent experience.

Session Memory

What Is Session Memory?

Session Memory is the information an AI Agent retains during a single interaction or conversation. This memory allows the agent to understand and maintain the context within that specific session, enabling more coherent and relevant responses.

How Does Session Memory Enhance Interactions?

  • Context Awareness: Remembers previous messages or actions within the session to provide contextually appropriate responses.
  • Dynamic Conversations: Enables back-and-forth dialogue that feels natural and human-like.
  • Task Continuity: Keeps track of ongoing tasks or inquiries without requiring you to repeat information.

Example of Session Memory in Action

Imagine you're chatting with an AI Agent about booking a flight:

  1. You: "I need to book a flight to New York next Monday."
  2. Agent: "Sure, do you have a preferred airline?"
  3. You: "No preference. Also, I might need a hotel."
  4. Agent: "Understood. I'll look for flights to New York next Monday and find hotel options for you."

The agent remembers that you're interested in flights to New York next Monday and that you might need a hotel, maintaining context throughout the conversation.

Long-Term Memory

What Is Long-Term Memory?

Long-Term Memory is the information an AI Agent retains over time, beyond individual sessions. This memory allows the agent to store learnings, preferences, and important information, enabling it to become more personalized and effective with each interaction.

How Does Long-Term Memory Enhance the Agent?

  • Learning from Experience: Remembers past interactions to improve future responses and actions.
  • Personalization: Stores user preferences and patterns to provide customized experiences.
  • Knowledge Accumulation: Builds a repository of information that the agent can draw upon to answer questions or perform tasks more efficiently.

Example of Long-Term Memory in Action

Continuing with the travel assistant example:

  • Past Interaction: You previously told the agent that you prefer window seats and like staying in boutique hotels.
  • Future Interaction: Weeks later, when you ask the agent to book another trip, it automatically selects window seats and suggests boutique hotels without you having to specify.

Interconnection Between Session and Long-Term Memory

The two types of memory work together to provide a cohesive experience:

  • Seamless Transitions: Information from session memory can be stored in long-term memory if it's deemed important for future interactions.
  • Enhanced Context: Long-term memory can inform session interactions by providing background information that enriches the conversation.
  • Continuous Improvement: The agent uses both memories to learn and adapt, becoming more effective over time.

How They Work Together

  1. During a Session: The agent uses session memory to maintain context and provide relevant responses.
  2. Post-Session: Important information from the session is stored in long-term memory.
  3. Future Sessions: The agent accesses long-term memory to personalize interactions and recall past preferences or important details.

Benefits of Memory in AI Agents

For Users

  • Personalized Experiences: Enjoy interactions tailored to your preferences and history.
  • Efficiency: Spend less time repeating information or clarifying requests.
  • Improved Accuracy: Receive more accurate and relevant responses as the agent learns over time.

For Organizations

  • Customer Satisfaction: Provide users with a more engaging and helpful experience.
  • Insights and Analytics: Gather valuable data on user interactions for continuous improvement.
  • Scalability: Leverage the agent's ability to learn and adapt without extensive reprogramming.

Best Practices

  • Regular Reviews: Periodically review what the agent has stored in long-term memory to ensure relevance and accuracy.
  • User Consent: Always inform users about the agent's memory capabilities and obtain necessary consents.
  • Security Measures: Make sure the agent complies with data privacy regulations and follows best practices for data security and retention on your region.