textmark-engine

In-memory full-text search for Node.js. Index documents, run TF-IDF ranked queries, boolean expressions, phrase search, prefix suggestions, and fuzzy matching.

Zero runtime dependencies. Node 18+.

Install

npm install textmark-engine

Quick start

import { createIndex } from "textmark-engine";

const search = createIndex({
  fields: ["title", "body"],
  boost: { title: 3 },
});

search.add({ id: "1", title: "Node streams", body: "Readable and writable streams in Node.js" });
search.add({ id: "2", title: "Search engines", body: "Build an inverted index with TF-IDF ranking" });
search.add({ id: "3", title: "Fuzzy matching", body: "Find terms even with small typos" });

search.search("node stream");
// [{ id: "1", score: ..., matched: [...], document: {...} }]

search.search('title AND "inverted index"');
search.search("engin", { prefix: true });
search.search("fuzzy", { fuzzy: 1 });
search.suggest("strea"); // ["stream"]

API

| Method | Purpose | | --- | --- | | createIndex(options) / new Textmark(options) | Create an engine | | add(doc) / addAll(docs) | Index documents | | update(id, patch) / remove(id) | Mutate the index | | search(query, options) | Ranked search | | suggest(prefix) | Autocomplete terms | | toJSON() / Textmark.fromJSON() | Snapshot / restore |

Search options

• limit — max hits (default 10)
• prefix — match term prefixes
• fuzzy — true (distance 1) or a number

Boolean query syntax

AND, OR, NOT, and "quoted phrases".

What it is for

Local search in CLIs, docs sites, admin tools, or tests — without Elasticsearch. It runs in one Node process.

License

MIT