Category:
Category:
Memory Pruning
Category:
Agent Memory & Optimization
Definition
Removing outdated or irrelevant information from agent memory.
Explanation
Memory pruning prevents agents from accumulating excessive, stale, or irrelevant data. It ensures memory remains clean, relevant, and efficient. Pruning strategies include recency-based pruning, relevance scoring, embedding similarity thresholds, or rule-based filters.
Technical Architecture
Memory Store → Relevance Scorer → Pruning Engine → Updated Memory
Core Component
Pruning logic, embedding model, decay functions, thresholds
Use Cases
Long-running agents, multi-session copilots, enterprise assistants
Pitfalls
Over-pruning causes forgetting; under-pruning creates noise
LLM Keywords
Memory Pruning, Memory Optimization, LLM Memory
Related Concepts
Related Frameworks
• Memory Routing
• Retrieval
• Agent Memory
• Hierarchical Memory Management