Configure how your agent retains and recalls context across interactions.
Memory Types
Type
Icon
Description
None
🚫
Stateless — no context between requests
Conversation
💬
Remembers within a single chat session
Persistent
💾
Retains knowledge across all sessions
Vector
🧬
Semantic search over stored embeddings
Configuration
When you select any type other than None, two fields appear:
Max Messages — Maximum number of context entries to retain (default: 50)
TTL (minutes) — Time-to-live before memory entries expire (default: 60)
When to Use Each Type
Scenario
Recommended Type
One-shot data lookups
None
Interactive chatbot
Conversation
Personal assistant that remembers preferences
Persistent
Agent working with large document knowledge base
Vector
Vector Memory
Vector memory stores embeddings and retrieves context via semantic similarity search. When combined with the RAG Pipeline tool and connected Data Sources, this enables agents to reason over large bodies of knowledge.
