PML Concepts

This section explains the core concepts of PML (Procedural Memory Layer).

Overview

PML is an intelligent MCP gateway that:

1. Routes requests to the right MCP tools
2. Learns from usage patterns
3. Suggests optimized workflows
4. Executes workflows with 0ms perceived latency via speculative execution

Key Metrics

Metric	Value	Impact
Context usage	<5%	90% de context window recuperee pour vos conversations
Latence workflows	5x → 1x	Parallelisation intelligente via DAG
MCP servers	15+ simultanes	Sans degradation de performance
Latence percue	0ms	Execution speculative pour les actions previsibles

Architecture 3 Couches

Reading Path

01 - Foundations

• 01 - MCP Protocol
• 02 - Gateway
• 03 - Database

02 - Discovery

• 01 - Semantic Search
• 02 - Hybrid Search
• 03 - Proactive Suggestions

03 - Learning

• 01 - GraphRAG
• 02 - Dependencies
• 03 - Confidence Levels
• 04 - Feedback Loop

04 - Capabilities

• 01 - What is a Capability
• 02 - Eager Learning
• 03 - Schema Inference

05 - DAG Execution

• 01 - DAG Structure
• 02 - DAG Suggester
• 03 - Parallelization
• 04 - Checkpoints
• 05 - Speculative Execution

06 - Code Execution

• 01 - Sandbox
• 02 - Worker Bridge
• 03 - Tracing

07 - Real-time

• 01 - Events
• 02 - Visualization