🧠 Memory Management in Lyzr Agents

Memory management in Lyzr is critical to building intelligent, context-aware agents. It allows agents to remember recent messages, retain useful information across sessions, and deliver coherent, personalized responses.

Memory Configuration Overview

Β Learn how to configure memory settings for your agents in Studio.

πŸ—‚οΈ Types of Memory

Lyzr supports two types of memory:

1. Short-Term Memory

Short-Term Memory stores recent messages from the current session. It enables the agent to remember context over multiple conversational turns (multi-turn dialog) without needing users to repeat themselves.
  • Session-scoped: Data exists only during a single conversation.
  • Limit-bound: You can control how many recent messages are retained using the slider.
  • Behavior:
    • Once the message limit is reached, older messages are summarized and moved to long-term memory (if enabled).
    • This ensures the agent still retains gist-level understanding of prior context.

Studio Configuration

You can configure the short-term memory window in Lyzr Studio:
πŸ“Œ Messages beyond this limit are summarized and stored as Long-Term Memory.

API Configuration

{
  "type": "SHORT_TERM_MEMORY",
  "config": {
    "max_messages": 16
  },
  "priority": 0
}

2. Long-Term Memory

Long-Term Memory persists important information across sessions, allowing the agent to recall what it learned from previous conversations.
  • Persistent: Stored in a secure backend store.
  • Summarized inputs: When short-term memory overflows, older data is automatically summarized and pushed to long-term memory.
  • Great for:
    • Remembering user preferences
    • Tracking ongoing projects or histories
    • Delivering personalized experiences

Studio Configuration

  • Enable Long-Term Memory in the Memory tab.
  • No manual storage limitsβ€”summarized memory grows as needed and can be queried programmatically.

API Configuration

{
  "type": "LONG_TERM_MEMORY",
  "config": {},
  "priority": 1
}

βœ… Summary: How They Work Together

FeatureShort-Term MemoryLong-Term Memory
ScopeCurrent session onlyAcross sessions
Retention controlConfigurable message windowSummarized data from ST memory
PersonalizationNoYes
Reset on session endYesNo
Use caseMulti-turn dialog coherencePersistent user knowledge
πŸ’‘ Lyzr agents automatically summarize and transfer older messages from short-term to long-term memory once the configured message limit is reached.

πŸ“˜ Next Steps

  • Configure memory per agent in Studio > Memory Tab.
  • Use the REST API if provisioning agents programmatically.
  • Monitor memory usage and debug context from the Agent Logs tab.
For advanced use cases, consider memory with custom priorities, vector store lookups, or external storage integration.