In this session, we present our experimental approach to creating DPK pipelines using agentic workflows. We will begin with a brief introduction to agentic workflows, followed by a walkthrough of two notebooks developed to support this work:

The first notebook shows a planner agent for Data-Prep-Kit tasks with code generation. The agent builds DPK pipeline that performs required tasks defined by a natural language.

The second notebook demonstrates how DPK transformers can be wrapped as tools within LangChain and LlamaIndex, along with examples of executing the transforms directly.