Learning

How PML learns from agent workflows.


Topics

Document Description
GraphRAG Knowledge graph structure and algorithms
Dependencies Tool and capability relationships
Confidence Levels template → inferred → observed
Feedback Loop How the system improves over time

Overview

PML learns in three ways:

  1. GraphRAG - Stores relationships between tools, patterns, and capabilities in a SuperHyperGraph (recursive structure supporting meta-capabilities)
  2. Dependency Tracking - Learns which tools work well together
  3. Confidence Evolution - Patterns gain confidence as they're validated through execution