Query Pipelines is a new declarative API that allows for orchestration of workflows within LlamaIndex to query over data. It is based on the QueryPipeline abstraction and offers advantages such as expressing common workflows with fewer lines of code, greater readability, and better integration with common low-code/no-code solutions. The API can be used in two main ways: as a sequential chain or as a full DAG, allowing for complex workflows. The post provides examples of how to use Query Pipelines for different scenarios, such as simple chains, query rewriting workflows, and full RAG pipelines. The post also references the Llamaindex documentation and an article by Jerry Liu for further information.
source update: Quickstart Guide to the Declarative… – Towards AI