Feature: Better Context Management via Core Memory

[actuallyrizzn]

Overview

Enhance The Librarian's context management capabilities by leveraging Letta's core memory system more effectively.

Goals

• Implement intelligent core memory updates based on conversation patterns
• Automatically extract and store important context in core memory
• Improve context retrieval and relevance for future conversations
• Reduce reliance on recall memory for frequently accessed information

Current State

Currently, The Librarian primarily uses recall memory for conversation history. Core memory is static and manually configured.

Proposed Enhancements

1. Automatic Core Memory Updates

   • Detect important information that should be persistent
   • Automatically update core memory with user preferences, key facts, and patterns
   • Implement heuristics for determining what belongs in core vs recall memory

2. Context Extraction

   • Extract key entities, relationships, and facts from conversations
   • Store structured information in core memory blocks
   • Enable semantic search across core memory

3. Memory Hierarchy

   • Establish clear hierarchy: Core Memory → Archival Memory → Recall Memory
   • Implement automatic promotion of important recall memory to core memory
   • Optimize memory access patterns

4. User Context Persistence

   • Store user-specific context in core memory
   • Enable per-user conversation continuity
   • Support multi-user scenarios with isolated contexts

Benefits

• Better long-term context retention
• Improved response quality through persistent knowledge
• Reduced token usage by storing context in memory rather than conversation history
• Enhanced personalization capabilities

Related

• Letta core memory API
• Archival memory system
• User identity management