Glossary

A

API Gateway

The Kong-based gateway that handles authentication, rate limiting, and request routing. All client traffic flows through the gateway before reaching the Core API.

API Key

Authentication credential for accessing the MemNexus API. Format: cmk_live_<id>.<secret>. Passed as a Bearer token in the Authorization header.

Artifact

A file or document associated with a memory or context. Artifacts can store supplementary data like PDFs, images, or configuration files.

C

Community

A group of related topics detected by graph algorithms (Louvain or label propagation). Communities represent natural clusters in your knowledge graph.

Conversation

A logical grouping of related memories. Represents a work session, debugging session, or research thread. MemNexus can auto-detect conversation boundaries using a 90-minute inactivity threshold.

Core API

The central Node.js/Express service that handles all data operations. Provides 40+ REST endpoints for memories, conversations, topics, facts, entities, patterns, and more.

Cosine Similarity

The metric used to compare vector embeddings. Measures the angle between two vectors, producing a score from -1 to 1 (in practice, 0 to 1 for embeddings). Higher scores indicate more similar meanings.

E

Embedding

A vector representation of text in high-dimensional space (1536 dimensions for MemNexus). Generated by OpenAI's text-embedding-3-small model. Used for semantic search.

Entity

A person, tool, project, or concept that appears across memories. MemNexus extracts and tracks entities to build a web of relationships.

Episodic Memory

A memory recording a specific event or experience tied to a time and context. The default and most common memory type.

F

Fact

A structured piece of knowledge stored as a subject-predicate-object triple (e.g., "MemNexus → built_with → Node.js"). Facts build an explicit knowledge graph for traversal and reasoning.

Full-Text Index

A Neo4j index powered by Apache Lucene that enables keyword-based search across memory content.

G

GraphRAG

Graph-based Retrieval Augmented Generation. Traverses the knowledge graph to find connections, communities, and relationships that provide richer context than flat search alone.

Graph Database

A database optimized for storing and querying relationships between data. MemNexus uses Neo4j.

H

Hybrid Search

The search approach that combines semantic (vector) and keyword (full-text) search, merging results using Reciprocal Rank Fusion.

I

Importance

A 0-1 score indicating how significant a memory is. Higher importance memories rank higher in search results. Recommended: 0.9 for decisions, 0.7 for context, 0.5 for general info.

K

Knowledge Graph

The graph of facts, entities, and their relationships. Built from explicit fact creation and automatic entity extraction.

L

Label Propagation

A community detection algorithm that iteratively assigns labels to nodes based on their neighbors' labels. One of two algorithms MemNexus supports (the other is Louvain).

Louvain Algorithm

A community detection algorithm that optimizes modularity to find dense clusters of connected nodes. Commonly used for topic community detection in MemNexus.

M

MCP (Model Context Protocol)

An open standard for connecting AI agents to external tools and data sources. MemNexus provides an MCP server with 19 tools for memory management.

Memory

The fundamental unit of storage in MemNexus. A discrete piece of information with content, topics, importance, timestamps, and a vector embedding.

N

Neo4j

The graph database used by MemNexus. Provides graph storage, vector indexes for semantic search, and full-text indexes for keyword search.

P

Pattern

A recurring theme or behavior detected from analyzing memories. Includes topic co-occurrence, temporal patterns, and behavioral trends.

Procedural Memory

A memory capturing a process, workflow, or step-by-step instructions (e.g., deployment procedures).

R

Reciprocal Rank Fusion (RRF)

The algorithm used to merge results from semantic and keyword search. Each result gets a score based on its rank in each search, producing a combined ranking.

S

Semantic Memory

A memory representing general knowledge or facts not tied to a specific event (e.g., "MemNexus uses Neo4j").

Semantic Search

Search that matches by meaning rather than keywords. Uses vector embeddings and cosine similarity to find conceptually related content.

T

Topic

A label applied to memories for categorization, clustering, and pattern detection. Topics can be discovered automatically or specified manually.

V

Vector Index

A Neo4j index that enables efficient nearest-neighbor search across vector embeddings. Used for the semantic component of hybrid search.