What it is
Typesense is a search engine written in C++ that focuses on instant search, typo tolerance, and (since v0.25) built-in vector search for semantic similarity. It was created in 2018 as an opinionated alternative to both Elasticsearch and Algolia and is now used by thousands of companies.
Why we recommend it
The killer feature is built-in vector search with hybrid keyword + semantic ranking. Most search engines force you to choose between keyword search (Elasticsearch, Meilisearch) and semantic search (dedicated vector DBs like Pinecone, Weaviate). Typesense does both with a single API. It's also written in C++ for raw performance — single-digit-millisecond search latency at scale.
When to use it
- Search that needs both keyword matching AND semantic understanding ("headphones" matches "earbuds" semantically)
- E-commerce sites with large catalogs
- Documentation sites with technical jargon
- Apps replacing Elasticsearch when they need search but not analytics
Quickstart
Run this in your terminal. Self-hosts locally with Docker or your preferred setup.
$ docker run -d -p 8108:8108 -v /tmp/typesense-data:/data typesense/typesense:27.1 Setup
Tags
Integrates with
Common use cases
- →product-search
- →docs-search
- →semantic-search
- →rag
- →ecommerce
Pricing
Self-host: free, Apache-2.0. Cloud: $0.0002 per search (PAYG) or $25/mo starter, $150/mo pro, $500/mo business.
Related projects
Apache Solr
by Apache Solr
Apache Solr is the original open source search server built on Apache Lucene. Battle-tested since 2004, full-text search, faceted search, geospatial, ML model upload, vector search (recent).
Manticore Search
by Manticore Search
Manticore Search is a fast, multi-threaded search engine. SQL-first (MySQL protocol), full-text search, vector search, faceted search, real-time indexing. The fastest ES alternative.
Meilisearch
by Meilisearch
Meilisearch is a search engine that you can drop into any app with a single Docker container. Sub-50ms search latency, typo tolerance, faceting, filtering, ranking rules, and a clean API. Free, MIT-licensed, no per-query fees.