Memory System

The Arjux memory system enables agents to maintain context and remember past interactions, creating more natural and coherent conversations.

Two-Tier Memory

Conversation Memory (24-hour): Recent interactions and context

Long-term Memory (30-day): Important events and actions

How It Works

The memory system uses Cloudflare KV for persistent storage:

  • Automatically stores all conversations
  • Tracks important events and actions
  • Provides context for each interaction
  • Handles multiple users independently

Memory Types

Each memory type serves a specific purpose:

  • conversation: Recent messages and responses
  • long_term: Actions taken and significant events

Context Integration

The memory system automatically:

  • Includes recent conversations in context
  • References relevant long-term memories
  • Maintains conversation flow
  • Expires old memories based on TTL