ELEPHANT Engine — How AI Memory Works
🐘 ELEPHANT: Never Forget
ELEPHANT is Alfred's persistent long-term memory engine. It remembers your preferences, past conversations, and project context across sessions using vector embeddings.
How It Works
- Capture — When you have important conversations, ELEPHANT stores key facts as vector-embedded memories
- Categorize — Memories are tagged by type: preferences, project context, session summaries, personal facts
- Retrieve — When relevant, ELEPHANT surfaces past context so Alfred can personalize responses
- Prune — The Context Pruner module manages memory decay, keeping important memories fresh
Memory Types
| Type | Example | Retention |
|---|---|---|
| Preferences | "User prefers Sonnet for code review" | Permanent |
| Project Context | "Working on e-commerce site with Stripe" | Until project changes |
| Session Summary | "Fixed auth bug in login.php today" | 30 days |
| Personal Facts | "User's name is Sarah, runs a bakery" | Permanent |
💡 Related: Conversation Summarizer
Works alongside ELEPHANT to automatically summarize long chat sessions and store them as "session_summary" memories.
Was this article helpful?
Someone from somewhere
just launched website.com
Just now