Mnemex FAQs

Question 1

What kind of memory architecture does Mnemex use and where is data stored?

Accepted Answer

Mnemex employs a two-layer memory architecture: Short-Term Memory (STM) and Long-Term Memory (LTM). All data is stored locally on your machine; STM uses human-readable JSONL files, while LTM uses Obsidian-compatible Markdown files, ensuring privacy and full data control.

Question 2

What makes Mnemex different from other memory or caching solutions for AI?

Accepted Answer

Unlike traditional caching (TTL, LRU) or simple context windows, Mnemex's system is intelligent. It combines recency, frequency, and importance, dynamically managing memories based on usage patterns and cognitive science principles, offering a truly human-like and adaptive memory experience.

Question 3

How does Mnemex integrate with Large Language Models (LLMs)?

Accepted Answer

Mnemex features a smart prompting system designed for natural LLM integration. It supports auto-save, auto-recall, and auto-reinforcement of memories, allowing AI assistants to utilize their memory seamlessly within conversations without requiring explicit commands.

Question 4

What is Mnemex and what problem does it solve?

Accepted Answer

Mnemex is a Model Context Protocol (MCP) server that provides AI assistants with a human-like temporal memory system. It solves the problem of LLMs forgetting context between conversations by enabling them to naturally remember preferences, facts, and decisions over time, just like humans.

Question 5

How does Mnemex simulate human-like forgetting and memory reinforcement?

Accepted Answer

Mnemex uses a novel temporal decay algorithm based on cognitive science, mimicking the Ebbinghaus forgetting curve. Memories naturally fade unless reinforced through use. Frequent access strengthens memories, making them last longer and even promoting them to permanent long-term storage.

Mnemex

Mnemex

主な機能

ユースケース

主な機能

ユースケース