Score Decay Visualization

How temporal weighting affects learning over time

Temporal Decay Factor

Recent outcomes matter more than old ones. The decay factor determines how quickly old data fades.

temporal_weight = decay_factor ^ days_since_execution
0.95
Today
1 day
3 days
7 days
14 days
30 days
60 days
90 days

Weighted Score Calculation

Example: A tool with mixed recent history

Days Ago Outcome Raw Value Weight Weighted
0 (today) Success 1.0 1.00 1.00
1 Success 1.0 0.95 0.95
3 Failure 0.0 0.86 0.00
7 Success 1.0 0.70 0.70
14 Success 1.0 0.49 0.49
Totals: 4.00 3.14
weighted_rate = 3.14 / 4.00 = 0.79 learning_boost = 0.5 + (0.79 * 1.5) = 1.68

Data Cleanup

🗑
M5-FIX-001: Data Cleanup Utility
Outcomes older than 90 days are automatically cleaned from both LanceDB (vectors) and Neo4j (graph). This keeps the learning system focused on recent, relevant data.