plugin-memory

@openlinkos/plugin-memory

Persistent memory plugin — part of the OpenLinkOS Agent Framework.

Overview

@openlinkos/plugin-memory gives agents long-term memory that persists across conversations. It supports vector storage for semantic retrieval, letting agents remember user preferences, past interactions, and learned knowledge.

Installation

pnpm add @openlinkos/plugin-memory @openlinkos/agent

Usage

import { createMemoryStore } from "@openlinkos/plugin-memory";

const memory = createMemoryStore({
  backend: "sqlite",
  maxEntries: 10000,
  retentionPolicy: "relevance",
});

// Store a memory
await memory.store("User prefers TypeScript over JavaScript", {
  category: "preferences",
  source: "conversation",
});

// Retrieve relevant memories
const results = await memory.retrieve("What programming language does the user like?");
console.log(results[0].content);
// → "User prefers TypeScript over JavaScript"

Features

• Vector storage — Store memories as embeddings for semantic search
• Semantic retrieval — Find relevant memories by meaning, not just keywords
• Multiple backends — In-memory, SQLite, and PostgreSQL storage options
• Retention policies — FIFO, LRU, or relevance-based entry eviction
• Metadata support — Tag memories with custom metadata for filtering

License

MIT