Vectra: a local vector database

Vectra is a local, file-backed, in-memory vector database with an optional gRPC server for cross-language access. Each index is a folder on disk — queries use MongoDB-style metadata filtering and cosine similarity ranking, with sub-millisecond latency for small indexes.

• Browser & Electron support — vectra/browser entry point with IndexedDBStorage and TransformersEmbeddings
• Local embeddings — LocalEmbeddings and TransformersEmbeddings run HuggingFace models with no API key
• Protocol Buffers — opt-in binary format, 40-50% smaller files
• gRPC server — vectra serve exposes 19 RPCs for cross-language access
• FolderWatcher — auto-sync directories into a document index
• Language bindings — vectra generate scaffolds clients for 6 languages

See the Changelog for breaking changes and migration details.

npm install vectra

import { LocalDocumentIndex, OpenAIEmbeddings } from 'vectra';

const docs = new LocalDocumentIndex({
  folderPath: './my-index',
  embeddings: new OpenAIEmbeddings({
    apiKey: process.env.OPENAI_API_KEY!,
    model: 'text-embedding-3-small',
    maxTokens: 8000,
  }),
});

if (!(await docs.isIndexCreated())) {
  await docs.createIndex({ version: 1 });
}

await docs.upsertDocument('doc://readme', 'Vectra is a local vector database...', 'md');

const results = await docs.queryDocuments('What is Vectra?', { maxDocuments: 5 });
if (results.length > 0) {
  const sections = await results[0].renderSections(2000, 1, true);
  console.log(sections[0].text);
}

Full docs at stevenic.github.io/vectra:

Vectra ships an llms.txt file that gives coding agents everything they need to integrate Vectra into your project. Point your agent at it and let it do the work:

Read the llms.txt file at https://raw.githubusercontent.com/Stevenic/vectra/main/llms.txt
and then add Vectra support to this project. Use LocalDocumentIndex for document
storage and retrieval.

The llms.txt file covers all exports, index types, CLI commands, gRPC bindings, and on-disk format — enough for any coding agent to scaffold a working integration without browsing docs.

MIT License. See LICENSE.

See CONTRIBUTING.md for guidelines. Please review our Code of Conduct.