@neuledge/graph-memory-registry

In-memory embedding registry for @neuledge/graph-router

Overview

@neuledge/graph-memory-registry provides an in-memory registry implementation that uses semantic similarity matching through embeddings. It's designed to work with @neuledge/graph-router to enable intelligent routing of queries to the appropriate data sources.

This package is part of the @neuledge/graph ecosystem. For general use with LLMs and AI agents, see the main package.

💻 Installation

pnpm add @neuledge/graph-memory-registry

npm install @neuledge/graph-memory-registry

yarn add @neuledge/graph-memory-registry

📖 Usage

Register data sources with embeddings and let the registry intelligently route queries:

import { NeuledgeGraphMemoryRegistry } from "@neuledge/graph-memory-registry";
import { openai } from "@ai-sdk/openai";

const registry = new NeuledgeGraphMemoryRegistry({
  model: openai.embedding("text-embedding-3-small"),

  // (optional) useful for persistence across restarts and serverless cold starts
  // commit this file to your repo or store in a shared volume
  cacheFile: "./embedding-cache.json", 
});

// Register a data source
registry.register({
  template: "weather.{city}",
  resolver: async ({ city }) => {
    // fetch weather for the matched city
    return { temperature: "72°F", condition: "Sunny" };
  },
});

// Match queries using semantic similarity
const match = await registry.match({ path: "weather.tokyo" });

🔧 API

• NeuledgeGraphMemoryRegistry - In-memory registry with embedding-based matching
  • register() - Register a data source with matcher and resolver
  • match() - Find matching data source for a query path
  • resolve() - Resolve a matched query to a concrete value
  • suggestions() - Get suggestions for alternative matches

For detailed API documentation, see the @neuledge/graph-router and main package.

📄 License

Apache-2.0 © Neuledge

See LICENSE for more information.