SuperLocalMemory Logo — Local AI Memory Layer
SuperLocalMemory
System Design

The V3 Architecture

Six retrieval channels. Three mathematical layers. One principle: your data, your machine, mathematically grounded.

Retrieval Engine

Six-Channel Hybrid Retrieval

No single retrieval method handles every query type. V3.3 runs six channels in parallel — including query completion where partial queries infer your full intent — and fuses them for maximum recall.

Semantic Channel

Fisher-Rao weighted embedding similarity. Models each memory as a probability distribution — not a flat vector. Graduated ramp from cosine to information-geometric distance over the first 10 accesses.

BM25 Channel

Classical keyword matching with persisted tokens. Handles exact names, rare terms, and technical identifiers that semantic similarity misses.

Entity Graph Channel

Spreading activation across the knowledge graph. 3-hop traversal with 0.7 decay. Finds relational connections between people, places, and concepts.

Temporal Channel

Date-aware retrieval with a 3-date model: authored, valid-from, valid-until. Handles questions like 'What was decided last Tuesday?'

Retrieval Pipeline

Query → Strategy Classification → 4 Parallel Channels

→ Weighted RRF Fusion (k=60)

→ Scene Expansion (pull all facts from matched scenes)

→ Bridge Discovery (multi-hop: Steiner tree + spreading activation)

→ Cross-Encoder Reranking (energy-weighted blending)

Top-K Results with per-channel scores

Mathematical Foundations

Three Novel Contributions

Each technique addresses a fundamental limitation in current memory systems. To our knowledge, none have been applied to agent memory before.

Fisher-Rao Geometry

Retrieval

The natural metric on statistical manifolds. Each memory embedding is modeled as a diagonal Gaussian with learned mean and variance. High-confidence memories score differently from uncertain ones — retrieval improves with use.

Sheaf Cohomology

Consistency

Algebraic topology for contradiction detection. The knowledge graph is modeled as a cellular sheaf. Computing H¹(G,F) reveals global inconsistencies that pairwise checking cannot detect — even when every local pair looks consistent.

Langevin Dynamics

Lifecycle

Memory lifecycle evolves via stochastic gradient flow on the Poincaré ball. The potential encodes access frequency, trust, and recency. Provable convergence to the stationary distribution — no hardcoded thresholds.

Full mathematical treatment with proofs and theorems in the V3 paper.

Read the Paper
Structured Encoding

11-Step Ingestion Pipeline

Every memory is processed through structured encoding before storage. This transforms raw text into a rich, queryable knowledge structure.

01

Metadata extraction — timestamps, source, importance

02

Entity resolution — canonical names with alias tracking

03

Fact extraction — atomic, typed facts (world / experience / opinion / temporal)

04

Knowledge graph construction — entities as nodes, relationships as edges

05

Temporal parsing — 3-date model (authored, valid-from, valid-until)

06

Emotional signal extraction — sentiment and emotional context

07

Scene clustering — group facts by temporal-semantic coherence

08

Observation building — structured entity profiles

09

Foresight generation — anticipatory indexing for future queries

10

Entropy gating — information-theoretic filtering (low-entropy = skip)

11

Compression and storage — write to 21-table SQLite schema

Operating Modes

You Choose the Privacy-Accuracy Tradeoff

A

Local Guardian

Zero cloud calls. All processing on your machine. EU AI Act compliant by architecture — data never leaves your device.

74.8%

LoCoMo (data stays local)

B

Smart Local

Mode A plus a local LLM via Ollama. Answer synthesis stays on your machine. Still fully private — nothing sent to any cloud.

Local LLM enhanced
C

Full Power

Cloud LLM for maximum accuracy. Cross-encoder reranking and agentic retrieval with multi-round refinement.

87.7%

LoCoMo (full power)

The Full Implementation is Open Source

Every algorithm, every test, every benchmark — available under AGPL v3.