Memory Consolidation
Master how AI agents consolidate short-term memories into efficient long-term knowledge bases
Your Progress
0 / 5 completedScoring Memory Importance
Not all memories are equally valuable. Importance scoring assigns weights to memories based on content quality, user signals, and contextual relevance to determine which deserve long-term storage.
Interactive: Importance Scoring Calculator
40%
30%
30%
Scored memories (threshold: 0.5):
Discussed transformers architecture
0.83Content: 0.9 • User: 0.7 • Context: 0.8
User is senior ML engineer
0.81Content: 0.9 • User: 0.8 • Context: 0.7
Prefers Python for data science
0.77Content: 0.8 • User: 0.9 • Context: 0.6
Said "hello"
0.10Clicked menu button
0.05⚖️ Scoring Factors
📝
Content Quality
Measures information density and uniqueness.
Facts > Preferences > Actions > Greetings
👤
User Signals
Explicit feedback: saves, bookmarks, corrections.
Saved = 1.0, Mentioned = 0.7, Passive = 0.3
🎯
Context Relevance
How often memory is retrieved and used.
Frequency × Recency × Utility
💡 Practical Strategies
•
Decay Over Time: Multiply scores by e^(-λt) so old memories fade unless repeatedly accessed.
•
Threshold Filtering: Only consolidate memories above threshold (e.g., 0.5). Rest are deleted.
•
Dynamic Weights: Learn optimal weights from user behavior over time using feedback loops.