bx-ai-memory

What problem does it solve?

BoxLang AI agents often need to remember context across turns, multiple users, and conversations. This skill provides typed memory abstractions (windowed, summary, session, file, cache, JDBC, vector) with multi-tenant isolation (userId + conversationId) and easy integration with agents and pipelines. It guides choosing the right memory type and how to apply it across scenarios.

Core Features & Use Cases

• aiMemory() supports multiple memory types (windowed, summary, session, file, cache, JDBC, chroma/pinecone/weaviate, in-memory-vector)
• Multi-tenant isolation using userId and conversationId to keep data segregated
• Vector memory options for semantic search and retrieval in RAG scenarios
• Easy integration with BoxLang agents and pipelines

Quick Start

Create a windowed memory with maxMessages 20 for userId 'user-alice' and attach it to an agent.