Add search that ranks with Elasticsearch, from inside your app or behind its own server: full-text, vector, hybrid, and geo on one typed index, in a single package. When you add AI, the same index backs every answer with the passage it came from.
Create an index with a typed schema, insert documents, and query with filters. The same engine works embedded in Node.js, inside the browser, or behind the bundled HTTP server.
The engine embeds in Node.js, Bun, Deno, and the browser, while the same package also serves a REST API from its own server. Multi-node cluster mode is experimental today, with TCP and gRPC transports between nodes.
Read the getting started guideimport { createNarsil } from '@delali/narsil'
const narsil = await createNarsil()
await narsil.createIndex('products', {
schema: {
title: 'string',
description: 'string',
price: 'number',
inStock: 'boolean',
},
language: 'english',
})
await narsil.insert('products', {
title: 'Mechanical Keyboard',
description: 'Cherry MX Brown switches with PBT keycaps',
price: 129.99,
inStock: true,
})
const results = await narsil.query('products', {
term: 'mechanical keyboard',
filters: {
fields: {
inStock: { eq: true },
price: { lte: 200 },
},
},
})The query that you write against an embedded index is the query that you send to the server, even after partitions shard the index behind it. You can move an index between deployments as one .nrsl snapshot, so when you scale, you change the deployment and keep your query code.
An index that you build in your process loads on the server unchanged, because the embedded engine and the server write the same binary format.
In your process
answers with no network hop
Behind the server
REST API with a write-ahead log
Across partitions
workers merge one ranked list
Narsil retrieves the passages that your model writes its answer from. Its highlighting then marks the matched text, so you can trace every claim in the answer back to a document. The server-app example shows this end to end, with a toggle that sends the same question through keyword, semantic, or hybrid retrieval so that you can watch the sources change.
The example numbers each citation and keeps the passage beside the answer.
Does vitamin D supplementation reduce respiratory infections?
The retrieved trials report a modest protective effect, strongest in participants who started with a deficiency and took daily doses rather than large single boluses.[1][2]
Sources
...daily vitamin D reduced the risk of acute respiratory infection among all participants...
...protective effects were strongest in those with a baseline deficiency...
Narsil loads its engine and index into the browser and persists the index to IndexedDB, so every search completes on the device. Web Workers keep the search off the main thread. The Cmd+K search on this site and on the main portfolio uses Narsil in this way.
The browser demo indexes a JSON or CSV file of your own inside the page, so the file stays on your machine.
Documentation
Hybrid search
Run keyword and vector retrieval in one query and fuse the rankings.
Vector search
Store embeddings in vector fields and search by similarity.
Pages
Benchmarks
Against Elasticsearch, OpenSearch, Qdrant, and Weaviate, with every engine on its recommended production settings and at the same search accuracy, Narsil processes 1,154 vector searches a second over 1M documents. The benchmark harness recorded every figure below on 28 September 2026.
1M
documents
1,154searches a second
Narsil processes the most vector searches a second over 1M documents.
Every engine searches the same DBpedia vectors on its recommended production settings and finds at least 99 per cent of the true nearest neighbours.
Keyword, vector, hybrid, and geo queries all search one typed index. The same package also includes partitioning and persistence.
Narsil is free and open source under the Apache 2.0 licence.