Core Concepts

Understand the building blocks of MemNexus — memories, conversations, topics, facts, and entities.

MemNexus organizes information into five core data types. Understanding these helps you use the platform effectively.

All five data types — memories, conversations, topics, facts, and entities — are connected in a graph. This is what powers relationship-based search and pattern detection.

Memories

A memory is the fundamental unit of storage. It represents a discrete piece of information — something that happened, a decision that was made, a preference you expressed, or context you want to preserve.

{
  "id": "2e61a084-776f-4bb0-9c7e-01ec2175d444",
  "content": "Deployed v1.26.2 of core-api. Fixed OpenAPI validation issues with nested schemas.",
  "topics": ["deployment", "core-api", "openapi"],
  "importance": 0.8,
  "timestamp": "2026-02-06T12:17:27Z"
}

Each memory has:

Content — The actual information being stored
Topics — Tags for categorization and filtering
Importance — A 0-1 score indicating how significant this memory is
Timestamps — When it was created and when the event occurred

Conversations

A conversation groups related memories into a logical unit of work. Think of it as a session — a debugging session, a feature implementation, or a research thread.

Conversations provide temporal context. Instead of searching across all memories, you can reconstruct what happened during a specific work session in chronological order.

MemNexus can automatically detect conversation boundaries using a 90-minute inactivity threshold, or you can explicitly create and manage conversations.

Topics

Topics are labels that emerge from your memories. They enable filtering, clustering, and pattern detection.

When you tag memories with topics like deployment, core-api, or debugging, MemNexus can:

Cluster related topics using co-occurrence analysis
Detect communities of related topics using graph algorithms
Show you which topics you work with most frequently

Facts

A fact is a structured piece of knowledge stored as a subject-predicate-object triple:

TypeScript  →  is_a  →  Programming Language
Neo4j       →  uses  →  Graph Storage
MemNexus    →  built_with  →  Node.js

Facts build an explicit knowledge graph that AI agents can traverse and reason over. Unlike memories (which are episodic and time-bound), facts represent persistent truths about your domain.

Entities

Entities are the people, tools, projects, and concepts that appear across your memories. MemNexus extracts and tracks entities to build a web of relationships.

When multiple memories mention "Neo4j", MemNexus recognizes it as a recurring entity and can show you all memories, facts, and topics related to it.

How They Connect

These concepts form a graph:

Memory ──[HAS_TOPIC]──→ Topic
Memory ──[MENTIONS]───→ Entity
Memory ──[BELONGS_TO]─→ Conversation
Fact   ──[ABOUT]──────→ Entity
Topic  ──[RELATED_TO]─→ Topic

This graph structure is what makes MemNexus powerful. Instead of flat keyword search, you can traverse relationships — finding all memories related to an entity, or discovering topics that frequently appear together.

Next Steps

Quick Start — Create your first memory
How It Works — Understand the architecture
Managing Memories — Full memory operations guide