How to configure Vector Embeddings & Intelligence Drivers

Enable vector embeddings, semantic search, and entity graph consolidation in MemoFS.

MemoFS is built with a deterministic, zero-key fallback (BM25 lexical search + rule-based entity parsing). When you're ready for semantic retrieval and automated knowledge graph extraction, you can plug in vector embedding and LLM intelligence drivers.

Available Intelligence Drivers

Driver	Package	Execution	Best For
BM25 Default	Built-in	100% Local / Zero Keys	Offline development, instant startup
Transformers.js	`@memofs/adapter-transformers`	100% Local / ONNX	Local semantic search with zero API costs
Voyage AI	`@memofs/adapter-voyage`	Cloud API	State-of-the-art code & technical recall
OpenAI	`@memofs/adapter-openai`	Cloud API	`text-embedding-3-small` / `text-embedding-3-large`
Cloudflare Workers AI	`@memofs/adapter-workers-ai`	Edge / Serverless	Serverless edge deployment with Cloudflare Workers

Option 1: Local Semantic Search with Transformers.js

Run vector embeddings completely on-device without sending data across the network:

Install the adapter:

npm install @memofs/adapter-transformers

Configure in .memofs/config.json:

{
  "recall": {
    "localEmbeddings": true,
    "embeddingModel": "Xenova/bge-small-en-v1.5"
  }
}

Option 2: Voyage AI Code Embeddings

Voyage AI offers domain-specific models tailored for code and technical repositories:

Install the adapter:
npm install @memofs/adapter-voyage

In your TypeScript setup:

import { MemoFS } from "@memofs/core";
import { createNodeFsMemoryStore } from "@memofs/core/node-fs";
import { createVoyageEmbedder } from "@memofs/adapter-voyage";

const memo = new MemoFS({
  store: createNodeFsMemoryStore({ rootDir: "." }),
  embedder: createVoyageEmbedder({
    apiKey: process.env.VOYAGE_API_KEY!,
    model: "voyage-code-3",
  }),
});

Option 3: OpenAI Embeddings

Use OpenAI's text-embedding-3-small or text-embedding-3-large:

Install the adapter:
npm install @memofs/adapter-openai

Initialize in code:

import { MemoFS } from "@memofs/core";
import { createNodeFsMemoryStore } from "@memofs/core/node-fs";
import { createOpenAIEmbedder } from "@memofs/adapter-openai";

const memo = new MemoFS({
  store: createNodeFsMemoryStore({ rootDir: "." }),
  embedder: createOpenAIEmbedder({
    apiKey: process.env.OPENAI_API_KEY!,
    model: "text-embedding-3-small",
  }),
});

Reindexing Memory

Whenever you switch embedding models, regenerate the vector indexes:

npx @memofs/cli index --rebuild