create-typesense-app

Typo-tolerant Typesense search, running locally, in one command.

Compatibility note: Requires Node.js 22+. Docker is optional. Without it, point at a Typesense you already run or a Typesense Cloud cluster.

Scaffolding your first search app

With NPM:

npm create typesense-app@latest

Or with npx:

npx create-typesense-app@latest

Then follow the prompts!

To skip them, pass a directory and --yes:

# npm 7+, extra double-dash is needed:
npm create typesense-app@latest my-search-app -- --yes

cd my-search-app
npm run dev

Type hary potr. It finds Harry Potter.

Options

-t, --template <id>   nextjs | qwik | solid | vanilla
    --dataset <name>  books (default) | empty
    --port <number>   host port for the local server (default 8108)
-y, --yes             take the defaults, no prompts
-h, --help            show this
-v, --version         print the version

Templates

id	Stack
nextjs	Next.js + react-instantsearch
qwik	Qwik + instantsearch.js, bundled by Vite
solid	SolidJS + instantsearch.js connectors, bundled by Vite
vanilla	Vanilla JS + instantsearch.js, bundled by Vite

Both give you the same app. A template is a manifest plus a file overlay, so adding one touches nothing else. See templates/README.md.

What you get

• A search box, hits with covers, and one facet
• Typesense in Docker, CORS already configured
• ~10,000 books indexed
• A schema and a seed script, short and commented, in the template's language
• A search-only API key in the browser, scoped to documents:search on one collection. The admin key stays server-side and never gets the public prefix.

Scripts

Script	Does
npm run seed	rebuilds the collection, and restores the search key if it went missing
npm run typesense:up	starts the server
npm run typesense:down	stops it, keeping your data