We’re excited to announce the all-new Sajari is in beta today. Sajari is a new proprietary search engine built from the ground up. With the release of this beta, we are introducing pipelines, a new and powerful way for app and ecommerce developers to configure search.
Our early access customers have seen incredible increases in search performance using pipelines; an ecommerce search customer has even seen a conversion rate increase of more than 10%! But before we dive into pipelines and the other features in this release, let’s talk about why and how we rebuilt search from the ground up.
Why build a new search engine?
Search engines have been around for a while now. The popular Apache Lucene project has powered most on-site search solutions for 21 years and provides the foundation for Solr and Elastic, two of the most popular search solutions today. There are newer search solutions too, including many point solutions built to solve vertical search problems, such as mobile search.
We set out to take a fresh approach with Sajari. We didn't want customers to have to choose between the power and flexibility of open source or speed of newer solutions. We built a search platform that allows engineers to deliver hyper-relevant search results at lightning fast speeds all while shortening the development cycle.
Sajari combines the speed and text matching capabilities of a traditional search engine with deep integration of vectors and newer index types to take full advantage of machine learning. It needed to work equally well for site, app, and ecommerce search.
Most search engines require a great deal of engineering to write custom rules, leverage real-time business data, and manage infrastructure to meet sudden spikes in queries. At the same time, most companies don’t have the luxury to hire a team of search engineers and SREs.
With Sajari, any developer can set up and configure search in minutes, and then optimize search results to meet the needs of the business. All without having to worry about infrastructure management. Developers can implement advanced search capabilities — like reinforcement learning — in minutes instead of weeks, and deliver advanced search experiences in days, not months.
Let’s dive into more detail to see how we can make that happen.
Introducing pipelines for search configurability
Pipelines are easily configurable YAML-based scripts that define a series of steps that are executed sequentially when indexing a record (record pipeline) or performing a query (query pipeline).
Pipelines can be versioned and tested in a real-time preview so you can tune your search to deliver fantastic results. A/B testing doesn’t cost extra computationally or financially — you don’t need to reindex your records. As mentioned in the introduction, our early access customer uses pipelines to continuously improve search performance. Improvement is at 10% and climbing.
There’s more in this release, too, including:
- Built-in machine learning: Implement reinforcement learning with a couple of lines of YAML and watch your results adapt to the behavior of your users.
- New APIs: All new APIs for data ingestion and processing.
- Cloud functions: Call out to cloud functions to enrich and transform your data.
If this sounds like something you might want to try, sign up for our beta today. The beta for app and ecommerce use cases works as a typical 14-day trial, but also gives you access to the new ecommerce and app search features. As a special incentive, through September we’re giving all beta testers 20k queries per month free (double our new free tier) for a year.
We can’t wait to see what you’ll build. And be sure to give us feedback on the beta on our new community.