MemoryIndexer 0.15.3

MemoryIndexer

Core abstractions and minimal implementations for LLM memory management.

This package provides the foundational interfaces, models, and lightweight implementations for the Memory Indexer system. It has minimal external dependencies, making it ideal for:

• Custom implementations of storage/embedding providers
• Unit testing with InMemory implementations
• Building your own memory management solution

Quick Start

services.AddMemoryIndexerCore(options =>
{
 options.Search.DefaultLimit = 10;
 options.VCM.WorkingMemoryCapacity = 7;
});

What's Included

Interfaces

• IMemoryStore - Memory storage operations
• IEmbeddingService - Embedding generation
• IScoringService - Memory relevance scoring
• ISessionStore - Session management
• IVirtualContextManager - Context window management

Models

• MemoryUnit - Core memory entity
• Session - Conversation session
• EntityTriple - Knowledge graph entities

Implementations (Minimal Dependencies)

• InMemoryMemoryStore - In-memory storage for testing
• MockEmbeddingService - Deterministic embeddings for testing
• DefaultScoringService - Hybrid scoring algorithm

For Full Features

Use MemoryIndexer.Sdk for production features:

• InMemory/SQLite storage (extensible via IMemoryStore)
• LMSupply embeddings (ONNX-based)
• MCP protocol tools
• OpenTelemetry observability

dotnet add package MemoryIndexer.Sdk

License

MIT