Memory System

Canoryn's local memory system allows workflows and assistants to store and retrieve persistent context.

Types of Memory

Unlike traditional chatbots that forget after each session, Canoryn assistants maintain persistent, local state:

┌─────────────────────────────────────────────┐
│              MEMORY LAYERS                  │
├─────────────────────────────────────────────┤
│  Short-Term    │ Current conversation       │
│  Long-Term     │ Facts, preferences, history│
│  Procedural    │ How to do things           │
└─────────────────────────────────────────────┘

Memory Types

Episodic Memory

Remembers specific events and interactions.

"Last Tuesday, you asked me to pause Spotify when Zoom opens."

Semantic Memory

Stores facts and general knowledge.

"Your preferred music genre is Lo-fi. Your work hours are 9-5."

Procedural Memory

Knows how to perform tasks (stored as blueprints).

"When you say 'focus mode', I should close Slack and enable DND."

Using Memory in Blueprints

Memory Query Node

Search the assistant's memory for relevant context:

[User Input] → [Memory Query] → [AI with Context] → [Response]

Memory Store Node

Explicitly save information to memory:

[User says "Remember..."] → [Memory Store] → [Confirmation]

Privacy

• All memory is stored locally on your Mac
• Encrypted at rest
• Never transmitted to external servers
• You can clear memory anytime in Settings

Best Practices

Interactions

1. Continuous Personalization: The more you interact with Canoryn, the more it aligns with your preferences.
2. Update Context: If you need to correct stored preferences, you can explicitly update them.
3. Review periodically: Check Memory Bank in Settings to see what's stored