AI Self-Learning Memory
What is Self-Learning Memory?
Self-learning memory is an AI feature that remembers your learning information across conversations. When you tell the AI Tutor that you are learning Python or struggling with loops, the AI stores this information and uses it in future conversations to personalize its responses.
How It Works
- After each AI conversation, the system extracts memorable information
- Information is categorized: preference, fact, skill, progress, goal
- Each memory is embedded as a vector for similarity search
- Memories are stored in the database using pgvector
- In new conversations, the AI retrieves relevant memories to add context
Memory Categories
| Type | Example |
|---|---|
| Preference | "Prefers Python over JavaScript" |
| Fact | "Completed the for-loop lesson" |
| Skill | "Strong in functions and lists, weak in OOP" |
| Progress | "Completed 3 of 10 lessons in Python Basics" |
| Goal | "Wants to become a web developer in 6 months" |
Managing Your Memories
You have full control over your AI memories.
View all memories
GET /ai/memoriesReturns a list of all memories the AI has recorded about you.
Delete a memory
DELETE /ai/memories/:idDeactivates a specific memory. It will no longer be used in future conversations.
Privacy
- Memories belong only to you and are not shared with other users
- Memories are stored securely in a Neon PostgreSQL database
- You can delete any memory at any time
- Under GDPR, you can request export or deletion of all your data
- Memories never include sensitive information like passwords or tokens
Self-learning memory requires at least the AI Starter tier ($5/month). See 4-tier AI system for details.
How AI Uses Memories
When you start a new conversation, the AI will:
- Analyze your question
- Search for relevant memories using vector similarity
- Add context from memories into the system prompt
- Respond based on both the question and memory context
For example: if the AI remembers that you are weak in object-oriented programming, when you ask about classes in Python, the AI will explain more slowly and use simpler examples.