What’s a query pipeline?
What’s a query pipeline?
This article explains what a query pipeline is, how it fits into your search implementation, and when to use it to optimize search for different audiences and use cases.
What is a query pipeline
A query pipeline is a set rules or model associations you can define to modify queries performed in Coveo-powered search interfaces. You can use query pipelines to apply different relevance tuning, filtering, and machine learning features based on which search interface sends the query. The features you define in a pipeline are applied to all queries before they’re sent to the index.
How a query pipeline fits in the system
When a user searches, the query passes through the pipeline where it can be transformed before reaching the index and before results are returned.
Every Coveo organization includes a default query pipeline that applies to all search interfaces unless you specify otherwise. Create additional pipelines using the Coveo Administration Console when your search interfaces serve different purposes or audiences.
Your customers and partners use your community portal to self-serve and find relevant information, while your support agents use an internal service portal to find relevant public and private items. You can create a separate pipeline for each portal, defining appropriate rules for each search interface's audience.
You can also use the query pipeline’s A/B testing feature to evaluate rule or model changes on a subset of users.
Why use query pipelines
Use query pipelines to customize search behavior for different audiences, contexts, or purposes.
Common use cases include:
-
Separate audiences with different needs: apply different rules or models for different search interfaces geared towards specific user groups.
-
Context-specific filtering: automatically filter results based on the search interface (for example, showing only public content in a customer portal while including private content in an internal portal).
-
Relevance tuning per use case: boost different types of content depending on whether users are searching for documentation, products, or support articles.
Query pipeline rules
Query pipeline rules can define the following:
-
-
Thesaurus entries: replaces or expands queries with synonyms.
-
Stop words: ignores unimportant words in queries.
-
-
-
Featured results: items appear at the top of search results when the query meets a specific condition.
-
Ranking expressions: modifies the order of results that match specified expressions and conditions.
-
-
Groups & campaigns: groups sets of result ranking rules that apply for specific contexts, audiences, or periods of time.
-
Advanced query pipeline rules.
-
Filters: adds hidden query expressions to all queries that go through the query pipeline.
-
Query parameters: overrides search parameter values for every query that goes through the query pipeline.
-
Ranking weights: determines the impact of ranking factors when establishing the order of results.
-
Triggers: establishes actions to be performed in the search interface following an event when a condition is met.
-
Query pipeline model associations
This section lists the models that can be associated with a query pipeline.
Machine learning models for knowledge sites
-
Automatic Relevance Tuning (ART): Learns from user search behavior to automatically boost the relevance of search results.
-
Content Recommendations (CR): Analyzes user browsing and search behavior to recommend content that’s most likely to be relevant to the current user.
-
Dynamic Navigation Experience (DNE): Learns from user behavior to order the most relevant facets and facet values at the top for a given query.
-
IPX Recommendations (IPXRECS): Recommends relevant content to users in a Coveo In-Product Experience (IPX) search interface, based on previous search and click activity.
-
Passage Retrieval (CPR): Uses embeddings and vector search to retrieve the most relevant passages from your indexed content to ground Retrieval-Augmented Generation (RAG) systems and large language models (LLMs).
-
Query Suggestions (QS): Recommends relevant queries to users as they type in the search box, based on past successful queries.
-
Relevance Generative Answering (RGA): Uses embeddings and vector search to retrieve the most relevant information from your indexed content and generates a grounded answer to a query.
-
Semantic Encoder (SE): Uses vector search to retrieve items from your indexed content based on semantic similarity with the query, improving relevance for complex queries and feeding results to RGA and CPR for answer generation and passage retrieval.
-
Smart Snippets: Displays a snippet of the most relevant search result item directly on the results page.
Machine learning models for commerce sites
-
Automatic Relevance Tuning (ART): Learns from commerce-specific user actions to improve the ranking of products presented in search results and on listing pages.
-
Dynamic Navigation Experience (DNE): Learns from user behavior to order the most relevant facets and facet values at the top for a given query.
-
Intent-Aware Product Ranking (IAPR): Ranks products on a search results page based on the visitor’s shopping intent.
-
Listing Page Optimizer (LPO): ranks products on Coveo-powered product listing pages (PLPs) by learning from shopper behavior.
-
Predictive Query Suggestion (PQS): Suggests queries to users as they type in a search box. suggest query candidates that are tailored to the current visitor’s shopping context.
-
Product Recommendation (PR): Suggests relevant products to end users navigating Coveo-powered commerce solutions.
-
Query Suggestions (QS): Recommends relevant queries to users as they type in the search box, based on past successful queries.
Order of execution
The following diagram illustrates the order of execution of query pipeline features:
What’s next?
Learn how to create and use a query pipeline.