Skip to content

Typesense 26.0

Updated: at 03:29 AM

I have been a frequent user of Typesense for the past three years across many of my projects. Setting it up has been a delight, and maintaining it has been relatively straightforward on BetaSeries, ComeUp, and CCEyes.

Started as a more cost-effective alternative to Algolia, Typesense quickly embraced AI technology last year by becoming one of the first major search databases to incorporate vector search. This feature proves especially valuable when paired with embedding models such as OpenAI embeddings.

The typical process of using embeddings with Typesense is quite straightforward:

  1. Create embeddings for the content you want to analyze.
  2. Add the embeddings to a vector database.
  3. Use a different set of vectors to query your database and locate similar entries, organized by score.

After several iterations of Typesense, the process became even simpler. The embedding part has now been integrated into the server, making it effortless to just send your data and have it automatically embedded with the desired model.

The most recent release of Typesense, which may be considered one of the largest version upgrades in open source history (jumping from 0.25 to 26.0, indicating that Typesense is now production-ready - a fact I can confirm), introduces several intriguing new AI features:

The Typesense RAG documentation is excellently crafted, featuring of course a TV series recommendation as an example.

Another significant time-saving feature for me is the JOIN feature, which allows for connecting one or more collections via common reference fields and joining them during query time. This feature employs a standard SQL paradigm that is now also available in Typesense.

If you have never tried Typesense and are in need of an easy and fast search database, consider downloading Typesense on your laptop today or trying out the Typesense cloud hosting service. You won’t regret it!