AbstractMemory Documentation

AbstractMemory is a multi-layer memory system that extends AbstractCore's BasicSession with sophisticated memory capabilities. It provides both voluntary memory operations (through exposed tools) and automated memory processes (background indexing, consolidation, and context reconstruction).

Documentation Structure

• Getting Started - Installation, setup, and first steps
• Architecture - System design and component overview
• Capabilities - What AbstractMemory can do
• Examples - Code examples and usage patterns
• Troubleshooting - Common issues and solutions

Quick Overview

AbstractMemory provides:

Voluntary Memory Operations (Tool-Based)

• remember_fact() - Explicitly store important information
• search_memories() - Query past interactions and stored knowledge
• reflect_on() - Analyze patterns across memories
• reconstruct_context() - Build rich context for current interaction

Automated Memory Processes

• Context Reconstruction - Automatically builds relevant context for each interaction
• Memory Indexing - Indexes all memories for semantic search
• Fact Extraction - Extracts knowledge from conversations in background
• Memory Consolidation - Periodically consolidates memories into core components

Dual Storage Architecture

• Markdown Files - Human-readable, version-controlled memory storage
• LanceDB Vectors - Semantic search with embeddings and metadata

Core Concepts

AbstractMemory organizes memory into layers similar to human memory systems:

• Core Memory - Identity components that emerge from interactions (purpose, values, personality, etc.)
• Working Memory - Current context, active tasks, and immediate focus
• Episodic Memory - Significant events, experiments, and discoveries
• Semantic Memory - Knowledge, concepts, and insights
• Library Memory - Documents and external knowledge sources
• User Profiles - Understanding of individual users through interaction patterns

Integration with AbstractCore

AbstractMemory is built on AbstractCore and leverages:

• BasicSession - Foundation for conversation management
• BasicExtractor - Semantic knowledge extraction
• EmbeddingManager - Vector embeddings for search
• StructuredOutputHandler - Response parsing and validation

Philosophy

AbstractMemory draws inspiration from human memory systems, providing:

1. Voluntary Control - Like conscious memory, you can explicitly choose what to remember and search
2. Automatic Processing - Like subconscious processes, memory indexing and consolidation happen automatically
3. Emergence - Identity components develop naturally from interaction patterns
4. Transparency - All memory operations are observable and debuggable
5. Dual Storage - Both human-readable files and machine-searchable vectors

The system aims to create more contextual, personalized AI interactions while maintaining full transparency and user control.