Tune search relevance in real time
Pipelines give you flexibility and granular control to customize results exactly the way you want. With pipelines, you can add powerful functionality to your search collections without complex coding or configuration.
Transform and enrich your search index
Configure data ingestion and search algorithms via simple YAML-based notation. This allows you to store your entire config-as-code. Pipelines give you flexibility to chain together series of steps to execute complex instructions to both define how your data is ingested and how to build and process your queries.
Do you want a pretty search solution or a highly configurable one? That’s how most search-as-a-service solutions are offered. We didn’t like that tradeoff. Pipelines give you the power to do just about anything — clone the config, audit tiny changes, view diffs between versions, or add powerful processing… all in just a few lines of markup.
Changing a setting is simple. How about triggering complex logic that can be composed to combine amazingly powerful search? If you love GitHub Actions you will love pipelines.
You want to call an external service. Go for it. Pipelines can call out to third-party services to process information. In practice this is used to turn images into colors and descriptions, classification, clustering, custom dynamic pricing. For example, you can call out to an external vision API to extract metadata from an image so you can filter on that data.
Every pipeline change creates a new version. Why? Because you can move forward and back in time at will. Every group of changes can be A/B tested and unwound as needed. You can even do a code diff inline if needed to see exactly what changed.
Design your search algorithm
Search algorithms are complicated, so we broke it down into a series of small, easy to understand steps that can be mixed and matched to design your perfect algorithm. Here are just a few of the things you can do with query pipelines:
Give users flexibility to filter search results and pinpoint the content they’re looking for using categories, pricing, or other fields.
Boost results with positive interactions such as clicks or cart adds, and decrease results for negative interactions. Or boost results based on margins or sales. All of this is managed through pipeline scoring with a few keystrokes.
Out of the box, most spelling, synonyms, and natural language processing is managed for you. But pipelines can be used to manage unique terms, spellings, or synonyms that may be specific to your business or industry.