Memory Configuration

Memory Configuration Overview

Configure persistent context management for AI agents with conversation continuity, multi-agent context sharing, and intelligent data retention across workflow executions.

Core Memory Capabilities

Conversation Continuity

Maintain conversation history across sessions with automatic token management and context preservation

Multi-Agent Context Sharing

Enable seamless context sharing between agents with hierarchical keys and access controls

Intelligent Data Management

Automated summarization, priority-based eviction, and token-aware capacity management

Privacy & Security

Built-in data redaction, compliance controls, encryption, and audit logging

Quick Start Configuration

Create Memory Resource

Define your memory configuration

memory/conversation.yaml

resource: memory
id: conversation_memory
description: Stores conversation history for agents
version: 1.0.0

type: token_based
max_tokens: 2000
max_context_ratio: 0.8

persistence:
  type: redis
  ttl: 24h

privacy_policy:
  redact_patterns:
    - '\b\d{3}-\d{2}-\d{4}\b'  # SSN pattern
  default_redaction_string: "[REDACTED]"

Configure Agent Memory

Connect agents to memory resources

agents/chat_agent.yaml

resource: agent
id: chat_agent

config:
  model: gpt-4
  temperature: 0.7

memory:
  - id: conversation_memory
    key: "user:{{.workflow.input.user_id}}"
    mode: read-write

Memory Operations

Use configured memory in workflows and agents with append, flush, and clear operations

Integration Patterns

Multi-agent memory sharing, context isolation, and collaborative workflows

Privacy & Security

Data redaction, compliance, encryption, and audit logging

Basic Memory Resource Setup

Foundation Configuration

Persistence Configuration

Memory Management Configuration

Privacy and Security Configuration

Data Redaction

Configure sensitive data protection:

privacy_policy:
  redact_patterns:
    # Personal identifiers
    - '\b\d{3}-\d{2}-\d{4}\b'  # SSN
    - '\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b'  # Email
    - '\b(?:\+?1[-.\s]?)?\(?[2-9]\d{2}\)?[-.\s]?\d{3}[-.\s]?\d{4}\b'  # Phone

    # Financial information
    - '\b\d{4}[\s-]?\d{4}[\s-]?\d{4}[\s-]?\d{3,6}\b'  # Credit cards

  redaction_config:
    default_redaction_string: "[REDACTED]"
    preserve_format: true
    case_sensitive: false

  non_persistable_message_types:
    - system
    - tool_internal
    - debug

Configuration Templates

Ready-to-Use Templates

Production-tested configuration templates for different deployment scenarios.

dev_memory.yaml

resource: memory
id: dev_memory
description: Development memory configuration
version: 1.0.0

key: "user:{{.workflow.input.user_id}}"
type: token_based
max_tokens: 2000

persistence:
  type: redis
  ttl: 8h

Use Case: Local development and testing
TTL: Short-term (8 hours)
Security: Basic (no sensitive data handling)

Configuration Reference

Core Configuration

Resource identification and basic settings:

resource: memory        # Resource type (always "memory")
id: unique_memory_id   # Unique identifier
description: "Purpose and context"
version: "1.0.0"       # Configuration version

# Memory key template
key: "user:{{.workflow.input.user_id}}"
# or hierarchical: "project:{{.project_id}}:user:{{.user_id}}"

# Memory type
type: token_based       # or message_count_based, buffer
max_tokens: 4000
max_context_ratio: 0.8
model: gpt-4

Available Template Variables:

{{.workflow.input.*}} - Workflow input parameters
{{.session_id}}, {{.project_id}}, {{.agent_id}}, {{.user_id}}

Eviction and Flushing

Control memory management strategies:

# Eviction policies
eviction_policy:
  type: fifo            # or lru, priority

# For priority eviction
eviction_policy:
  type: priority
  priority_keywords: ["error", "critical", "important"]

# Flushing strategies
flushing_strategy:
  type: hybrid_summary  # or simple_fifo, lru, token_aware_lru
  summarize_threshold: 0.8
  summary_tokens: 500

Persistence and TTL

Configure data storage and expiration:

# Redis persistence (recommended)
persistence:
  type: redis
  ttl: 24h
  circuit_breaker:
    enabled: true
    timeout: "100ms"
    max_failures: 5

# TTL fine-tuning
append_ttl: "30m"    # Extend on append
clear_ttl: "5m"      # Short cleanup
flush_ttl: "1h"      # Preserve after flush

TTL Formats: 30s, 5m, 2h, 24h, 168h

Token Providers

Configure accurate token counting:

# OpenAI provider
token_provider:
  provider: openai
  model: gpt-4
  api_key_env: OPENAI_API_KEY
  fallback: tiktoken

# Anthropic provider
token_provider:
  provider: anthropic
  model: claude-3-sonnet
  api_key_env: ANTHROPIC_API_KEY

# Local fallback
token_provider:
  provider: tiktoken
  model: gpt-4

Privacy and Security Reference

Protect sensitive information:

privacy_policy:
  redact_patterns:
    - '\b\d{3}-\d{2}-\d{4}\b'  # SSN
    - '\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b'  # Email
    - '\b\d{4}[\s-]?\d{4}[\s-]?\d{4}[\s-]?\d{3,6}\b'  # Credit cards

  non_persistable_message_types:
    - system
    - tool_internal
    - debug

  default_redaction_string: "[REDACTED]"

Environment Variables

Use environment variables for sensitive data:

# Direct reference
api_key: "{{ .env.OPENAI_API_KEY }}"

# With fallback
api_key: "{{ .env.OPENAI_API_KEY | default \"fallback-key\" }}"

# Conditional
api_key: "{{ if .env.USE_OPENAI }}{{ .env.OPENAI_API_KEY }}{{ else }}{{ .env.FALLBACK_KEY }}{{ end }}"

Memory Configuration

Conversation Continuity

Multi-Agent Context Sharing

Intelligent Data Management

Privacy & Security

Create Memory Resource

Configure Agent Memory

Memory Types

Memory Keys

Complete Configuration Template

Redis Persistence (Recommended)

TTL Management

Eviction Policies

Flushing Strategies

On this page