Typesense

Typesense

★ 26,531971 forks971 contributorsUpdated 9/1/2026

Typesense is an open-source, typo-tolerant search engine built for fast instant-search experiences, and a self-hosted alternative to Algolia. It holds indexes in memory for sub-50ms queries, ships with sensible relevance defaults, and adds faceting, filtering, geo-search, and vector/semantic search. Official clients and framework adapters (InstantSearch, Next.js, Laravel) make integration quick.


Overview

Description

Typesense is an open-source, typo-tolerant search engine optimized for instant "search-as-you-type" experiences, and a self-hosted alternative to Algolia. It keeps indexes in memory to return results in milliseconds and is designed for a gentle learning curve — good relevance defaults, a simple REST API, and few knobs to tune. Alongside keyword search it supports vector and hybrid semantic search.

Features

  • Typo tolerance: Configurable fuzzy matching that corrects misspellings automatically.
  • Fast faceting & filtering: Real-time facet counts, filtering, and sorting over large datasets.
  • Semantic & hybrid search: Built-in vector search and rerankers, with optional LLM embeddings.
  • Geo search: Radius and bounding-box queries with distance sorting.
  • High availability: Raft-based clustering for replicated, fault-tolerant deployments.
  • Ecosystem: Official SDKs plus InstantSearch, Next.js, Laravel Scout, and Firebase adapters.

Technology Stack

  • Single self-contained C++ binary (also distributed as a Docker image)
  • RocksDB for on-disk persistence; indexes served from RAM
  • Raft consensus for clustering
  • REST/HTTP API with official client libraries

Requirements

  • Linux host or Docker
  • RAM roughly proportional to index size (the dataset must fit in memory)
  • A persistent data directory and an API key for bootstrapping

Learning resources