Nota Memoria Engine (NME)

NME is a memory structuring service that extracts and organizes atomic units of meaning (traits) from raw input before storage in Resonant Field Storage (RFS). Inspired by DeepMind's Titans memory architecture, NME transforms unstructured data into structured trait-based memories. It implements episodic, semantic, working, and long-term memory patterns, extracting traits like intent, entities, time, sentiment, domain, role context, constraints, priority, dependency, and result. NME ensures structured memories can be deterministically encoded into RFS waveforms.

DeepMind Titans Inspiration

• Episodic Memory: Temporal trait organization with time as first-class dimension
• Semantic Memory: Trait abstraction with abstracted knowledge (not raw events)
• Working Memory: Active trait context for current task
• Long-Term Memory: Persistent trait storage with retrieval mechanisms

Trait Extraction

NME extracts and structures atomic units of meaning (traits) including: intent, entities, time, sentiment, domain, role context, constraints, priority, dependency, and result. Each trait is an atomic unit of meaning that enables structured memory encoding.

Architectural Stack

• Trait Extraction Layer: Extracts atomic units of meaning with accuracy ≥ 0.95
• Memory Structuring Layer: Organizes traits into DeepMind Titans patterns
• Encoding Integration Layer: Encodes structured memories into RFS waveforms
• Pattern Classification Layer: Classifies memories with accuracy ≥ 0.90

KPIs & Guardrails

• Trait Extraction Accuracy: ≥ 0.95
• Trait Coverage: ≥ 0.85
• Structure Validation Pass Rate: = 1.0
• Pattern Classification Accuracy: ≥ 0.90
• Encoding Reconstruction Error: ≤ 1e-6
• Trait Extraction Latency (P95): ≤ 50 ms

For detailed technical documentation, see the NotaMemoriaEngine repository documentation.