HTM User Guides

Welcome to the HTM (Hierarchical Temporal Memory) user guide collection. These guides will help you understand and effectively use HTM for building intelligent LLM-based applications with persistent memory.

What is HTM?

HTM is an intelligent memory management system for LLM robots that implements a two-tier architecture:

• Working Memory: Token-limited active context for immediate LLM use
• Long-term Memory: Durable PostgreSQL storage for permanent knowledge

HTM enables your robots to recall context from past conversations using RAG (Retrieval-Augmented Generation), creating continuity across sessions and enabling sophisticated multi-robot collaboration.

Guide Categories

Getting Started

Perfect for developers new to HTM or those building their first application.

• Getting Started Guide - Your first HTM application, basic concepts, and common patterns

Core Operations

Learn how to work with HTM's memory system effectively.

• Adding Memories - How to store different types of information in HTM
• Recalling Memories - Search strategies and retrieval techniques
• Working Memory Management - Understanding token limits and eviction
• Long-term Memory - Database operations and maintenance

Advanced Features

Dive deeper into HTM's powerful capabilities.

• Multi-Robot Usage - Building hive mind systems with multiple robots
• Search Strategies - Vector, full-text, and hybrid search
• Context Assembly - Creating optimized context for LLMs

Integrations

Connect HTM to external tools and services.

• MCP Server - Use HTM with Claude Desktop, Claude Code, and AIA

Learning Path

We recommend the following progression:

1. Start Here: Getting Started Guide
2. Understand HTM's architecture
3. Build your first application

4. Learn basic operations

5. Core Skills: Memory Operations

6. Adding Memories - Store information effectively
7. Recalling Memories - Retrieve what you need

8. Context Assembly - Use memories with LLMs

9. Deep Understanding: Memory Management

10. Working Memory - Token management
11. Long-term Memory - Database operations

12. Search Strategies - Optimize retrieval

13. Advanced Topics: Multi-Robot Systems

14. Multi-Robot Usage - Build collaborative systems

Quick Reference

Common Tasks

• Initialize HTM: See Getting Started
• Add a memory: See Adding Memories
• Search for memories: See Recalling Memories
• Create LLM context: See Context Assembly
• Monitor memory usage: See Working Memory
• Multi-robot setup: See Multi-Robot Usage
• Use with Claude/AIA: See MCP Server

Memory Types

HTM supports six memory types, each optimized for different use cases:

Type	Purpose	Example
:fact	Immutable facts	"User's name is Alice"
:context	Conversation state	"Discussing database architecture"
:code	Code snippets	"Ruby function for parsing dates"
:preference	User preferences	"Prefers dark theme"
:decision	Design decisions	"Chose PostgreSQL for storage"
:question	Unresolved questions	"Should we add caching?"

Search Strategies

Strategy	Method	Best For
Vector	Semantic similarity	Conceptual searches, related topics
Full-text	Keyword matching	Exact terms, specific phrases
Hybrid	Combined approach	Best overall accuracy

Getting Help

• Examples: Check the examples/ directory in the HTM repository
• API Reference: See the API documentation
• Tests: Look at test/ directory for usage examples
• Issues: Report bugs on GitHub

Documentation Conventions

Throughout these guides, you'll see these admonitions:

Tip

Helpful advice and best practices

Warning

Important warnings about potential issues

Note

Additional information and context

Example

Code examples and usage demonstrations

Next Steps

Ready to get started? Head over to the Getting Started Guide to build your first HTM application.