How much RAM and disk does Typesense need on a VPS?

Typesense loads its indexes into memory. A 1 GB VPS is enough to index up to a few hundred thousand documents. For larger corpora or vector fields (embeddings), plan for 4–8 GB of RAM. Disk usage mirrors memory: a 300 k document index typically takes 200–500 MB. The ServOrbit client area shows live RAM usage; upgrade the plan at any time without data loss.

Does Typesense support Arabic, French and multilingual search?

Yes. Typesense ships a multilingual tokeniser that handles Arabic, French, English and 15+ other languages. Arabic right-to-left text and French accented characters (é, à, ç) are handled correctly without any plugin. For a bilingual Arabic/French catalogue, simply send both language fields in the same document and query across them.

Kick off a software project quickly.

Typesense

Typesense — C++ search engine: typo-tolerant, faceted, and vector search via REST API. Single container, under 150 MB RAM, GPLv3 licence.

1 GB (4 GB for large corpora or vector search) RAM 1 vCPU Port 8108 Available

Tech stack

DockerC++

Minimum RAM1 GB (4 GB for large corpora or vector search)

Minimum CPU1 vCPU

Default port8108

Compatible OSChoice of Linux distributions

Typesense (GPLv3, 26 k+ GitHub stars, v29.0) is an open-source search engine written in C++. It exposes a clean REST API for indexing and querying your data with typo tolerance, faceted filtering, vector/semantic search, and millisecond response times — all in a single Docker container with no JVM, no Elasticsearch cluster setup, no shard tuning.

Unlike Meilisearch (its closest peer), Typesense is optimised for lower latency at higher concurrency and supports multi-vector search natively, making it well-suited for hybrid keyword-plus-embedding search pipelines. The two engines coexist in many architectures: Meilisearch for its ease of use on smaller datasets, Typesense for strict p99 latency requirements or semantic ranking.

Deployed on a ServOrbit VPS, Typesense becomes the search layer of your application — whether you are building a multilingual product catalogue, a documentation portal with instant search, or a semantic search feature powered by embeddings from OpenAI or a local model. Official client libraries cover JavaScript/TypeScript, Python, PHP, Ruby, Go, Dart, Java and Swift.

The REST API is compatible with the Instantsearch.js / react-instantsearch family, so adding a polished, Algolia-like search UI to your front-end requires only a few lines of configuration.

Key features

Typo tolerance with fuzzy matching — a query for 'Mohamad' finds 'Mohamed' without any dictionary or configuration

Faceted search and filtering — filter by category, price, tag or any field alongside the full-text query, with live counts

Built-in vector / semantic search — index embedding vectors alongside text fields and run hybrid keyword + ANN queries natively

C++ engine with in-memory indexes and RAFT persistence — no JVM, no garbage collection pauses, designed for strict SLA requirements

Scoped API keys — generate search-only, index-specific or rate-limited keys to secure front-end bundles without exposing the admin key

GPLv3 licence, 26 k+ GitHub stars, actively maintained — weekly releases, official SDKs for 9 languages

When to use this solution?

1

Product catalogue search

Index your products with price, category and stock attributes. Users get instant, typo-tolerant results as they type, with facets to narrow by brand or price range — all from your VPS, with no per-query fee and no data leaving your infrastructure.

2

Documentation and knowledge base instant search

Crawl your Markdown docs, Ghost blog or Docusaurus site into Typesense and expose a search-as-you-type bar. The multilingual tokeniser handles Arabic, French and English natively, making it ideal for bilingual MENA-targeted documentation.

3

Hybrid keyword + semantic search

Store OpenAI or local-model embeddings alongside your text fields and run hybrid queries that combine BM25 keyword ranking with ANN vector similarity — all without a separate vector database. Typesense merges both scores with a single API call.

Deploy Typesense on your VPS

Guide optimized for ServOrbit Cloud VPS.

01

Order the ServOrbit VPS

Typesense idles under 150 MB RAM for typical corpora; a 1 GB VPS is enough to get started. Provisioning installs Docker on the Ubuntu image and prepares the stack.

02

Deploy from the Marketplace

From your ServOrbit client area, go to Marketplace → Development → Typesense → Deploy. A single typesense/typesense:29.0 container starts with a named volume for /data and a randomly generated admin API key shown in your client area.

03

Sign in for the first time

Typesense has no login screen: it is an API-driven search engine protected by an API key. Use the key provided in your client area as the X-TYPESENSE-API-KEY header on every request, for example: curl -H 'X-TYPESENSE-API-KEY: <your-key>' http://127.0.0.1:8108/health

04

Reach the API from your applications

The REST API listens on port 8108, published on the loopback of the VPS. Expose it via an Nginx or Caddy vhost on your domain over HTTPS for remote applications, or use an SSH tunnel for admin sessions: ssh -L 8108:127.0.0.1:8108 root@<your-vps-ip>, then http://localhost:8108.

05

Index your first data and query it

Create a collection and add documents: curl -X POST 'http://127.0.0.1:8108/collections' -H 'X-TYPESENSE-API-KEY: <key>' -H 'Content-Type: application/json' -d '{"name":"products","fields":[{"name":"name","type":"string"},{"name":"price","type":"float"}]}'. Then search: curl 'http://127.0.0.1:8108/collections/products/documents/search?q=vps&query_by=name' -H 'X-TYPESENSE-API-KEY: <key>'.

Frequently asked questions

Both are self-hosted, open-source search engines with typo tolerance and a REST API. Typesense (C++, GPLv3) is optimised for strict low-latency requirements and includes native vector / semantic search. Meilisearch (Rust, MIT) has a simpler setup and is often preferred for smaller datasets or teams that want zero configuration. If you already use Meilisearch, you can run Typesense alongside it for workloads that need hybrid keyword+vector ranking or strict low-latency requirements.

Yes. Using the admin API key, you can create scoped keys with fine-grained permissions: documents:search only on a specific collection, with optional embedded filters that cannot be overridden by the client. Expose this search-only key in your JavaScript bundle; keep the admin key server-side only.

Yes. Typesense publishes its API on 127.0.0.1:8108 and can share the VPS with your Laravel, Node.js, Django or any other application. Each application can call the local API directly without crossing the network. The ServOrbit provisioner allocates a port in the 20 000–29 999 range if 8108 is already taken, and the client area always shows the effective port.

Embed the deploy button

Maintaining a project that uses Typesense? This button lets your readers deploy it on a VPS in one click, without reading Docker documentation.

Markdown

[![Deploy Typesense on ServOrbit](https://servorbit.com/brand/deploy/button.svg)](https://servorbit.com/vps-cloud?template=typesense&utm_source=deploy-badge&utm_medium=referral&utm_campaign=typesense)

HTML

<a href="https://servorbit.com/vps-cloud?template=typesense&utm_source=deploy-badge&utm_medium=referral&utm_campaign=typesense"><img src="https://servorbit.com/brand/deploy/button.svg" alt="Deploy Typesense on ServOrbit" height="40"></a>

The button points to a VPS order with the template preselected. The image is served from servorbit.com — nothing to host on your side.

Complementary solutions

Development

Meilisearch

Self-hosted full-text search engine: add instant, typo-tolerant search to your applications with a simple REST API — 57 k GitHub stars, single Rust container, under 100 MB RAM.