n8n-nodes-token-aware-memory

This is an n8n community node that provides token-aware memory management for AI workflows with Redis persistence and automatic compression.

The Token-Aware Memory node stores conversation history with intelligent token management, hierarchical memory organization, and automatic summarization when token limits are approached.

n8n is a fair-code licensed workflow automation platform.

Installation
Operations
Configuration
Compatibility
Usage
Resources
Version history

Installation

Follow the installation guide in the n8n community nodes documentation.

Operations

Memory Management

• Store Messages: Automatically stores every user and AI message sent through connected nodes
• Token Monitoring: Tracks total token usage in real-time
• Hierarchical Storage: Organizes memory into short-term, mid-term, and long-term levels
• Automatic Compression: Compresses older messages when reaching 80% of maxTokens limit
• Full History Retrieval: Returns complete conversation history when requested

Memory Levels

• Short-Term: Recent messages stored verbatim
• Mid-Term: Partially summarized older messages
• Long-Term: Fully compressed historical summaries

Configuration

Required Parameters

• Max Tokens: Maximum total tokens allowed before triggering compression (default: 8000)
• Redis URL: Redis connection URL in format redis://[password@]host:port[/database] (default: redis://localhost:6379)
• Summarization Prompt: Custom prompt template for LLM-based compression

Optional Parameters

• Session ID: Unique session identifier to separate memory between different conversations/executions (leave empty for auto-generated)

Connections

• AI Language Model Input: Connect an LLM node for intelligent message summarization during compression

Compatibility

• Minimum n8n version: 1.0.0
• Requires Redis server for memory persistence
• Tested with Redis 6.0+
• Note: Redis dependency may not be compatible with n8n Cloud deployments. For cloud usage, consider alternative memory solutions or contact n8n support.

Usage

Basic Setup

1. Add the Token-Aware Memory node to your workflow
2. Configure Redis connection parameters
3. Connect AI memory output to nodes that need conversation history
4. Optionally connect an LLM node for summarization

Memory Persistence

Memory is automatically persisted to Redis and survives workflow restarts. Each workflow and node instance maintains separate memory spaces.

Token Management

• Messages are automatically compressed when total tokens reach 80% of maxTokens
• Compression uses connected LLM for intelligent summarization
• Fallback to simple truncation if no LLM is connected

Example Workflow

[AI Chat Node] → [Token-Aware Memory] → [AI Response Node]
                     ↓
              [LLM Node for Summarization]

Redis URL Examples

• redis://localhost:6379 - Local Redis without password
• redis://:password@localhost:6379 - Local Redis with password
• redis://:password@remote-host.com:6379/1 - Remote Redis with password and database 1

Session Isolation

Each Token-Aware Memory node instance uses isolated Redis keys based on:

• Workflow ID
• Node ID
• Session ID (user-provided or auto-generated)

This ensures that multiple conversations or workflow executions don't interfere with each other's memory. Use the Session ID parameter to manually control session grouping.

Resources

• n8n community nodes documentation
• Redis documentation

Version history

0.1.3

• Initial release with hierarchical memory management
• Redis persistence support
• Token-aware automatic compression
• LLM integration for summarization