Memorandum FAQs

Question 1

What types of memory does Memorandum manage?

Accepted Answer

Memorandum manages two core memory types: 'Facts' – short key-value records with LRU eviction and optional TTL, and 'Documents' – structured storage supporting YAML indexes, Markdown frontmatter, metadata, and binary files, all enhanced with semantic search capabilities.

Question 2

How does Memorandum enable semantic search for AI agents?

Accepted Answer

It leverages local HuggingFace embedding models (e.g., multilingual-e5-small) to perform semantic search across both facts and documents. This allows AI agents to retrieve information based on meaning and relevance, rather than just keywords.

Question 3

What is Memorandum and what is its primary purpose?

Accepted Answer

Memorandum is a software tool designed to provide AI agents with persistent memory and comprehensive knowledge management capabilities. It stores both short-term facts and long-term structured documents to enhance an AI's operational intelligence and recall.

Question 4

Can Memorandum be integrated into existing AI development environments?

Accepted Answer

Yes, Memorandum functions as an MCP server and offers straightforward integration. Configuration examples are provided for popular AI development tools like Claude Desktop and Claude Code, making it easy to set up and use with your AI projects.

Question 5

What measures does Memorandum take to ensure data durability and safety?

Accepted Answer

Memorandum prioritizes data durability through atomic file writes (using temporary files and renames to prevent corruption) and configurable autosave intervals. This ensures crash safety and prevents data loss for your AI agent's critical memory.

Memorandum

Memorandum

主要功能

使用案例

主要功能

使用案例