Memory Context Window Optimization
🎯 Overview
SuperOptiX provides GEPA-based context window optimization for memory systems, intelligently selecting which memories to include in the agent's context. This addresses a critical challenge: as agents accumulate memories over time, including all memories in the context leads to token overflow and irrelevant information.
Key Innovation: GEPA learns to select only the most relevant memories within your token budget.
Impact: - Token usage: 60% reduction (5000 → 2000 tokens) - Memory relevance: 55% improvement (30% → 85%) - Task success rate: 30-50% boost
⚡ Quick Start
Enable in Playbook
spec:
memory:
enabled: true
enable_context_optimization: true # Enable GEPA optimization
max_context_tokens: 2000 # Set token budget
Use in Agent
# Pull demo agent
super agent pull customer_support_memory
# Compile and run
super agent compile customer_support_memory
super agent run customer_support_memory \
--customer_query "What happened with my shipping issue?"
The agent automatically uses optimized context!
🔍 The Problem
Without Optimization
After 20+ interactions, agents accumulate many memories:
``` Query: "What happened with my shipping issue?"
Unoptimized Context (ALL 20 memories): 1. Order #AAA placed (Sept 1) - 200 tokens 2. Order #BBB placed (Sept 5) - 200 tokens 3. Order #CCC placed (Sept 10) - 200 tokens ... 15. More old orders - 200 tokens each 16. Shipping issue with #12345 (Oct 18) - 300 tokens ← RELEVANT! 17-20. More irrelevant data - 800 tokens
Total: 5000+ tokens → Context overflow! Relevant: 300 / 5000 = 6% ```
Problems: - Context overflowHuman: Can you also now ensure you add the entry to navigation