Pipeline Inspection, Visualization, and Interoperability in PyTerrier

TL;DR

PyTerrier enables declarative information retrieval pipelines by composing transformers; however, understanding, visualizing, and interoperating these pipelines remains challenging. The paper demonstrates three enhancements: an inspection module for programmatic input/output specifications and subtransformers, interactive schematics for visualizing data flow in notebooks and docs, and a Model Context Protocol (MCP) server to expose pipelines to external tools and agents. These contributions enable automatic pipeline validation, richer visualization, and seamless integration with LLM-assisted workflows and tools like Copilot. The work has practical impact by lowering barriers to constructing, understanding, and deploying IR pipelines across research and applied settings.

Abstract

PyTerrier provides a declarative framework for building and experimenting with Information Retrieval (IR) pipelines. In this demonstration, we highlight several recent pipeline operations that improve their ability to be programmatically inspected, visualized, and integrated with other tools (via the Model Context Protocol, MCP). These capabilities aim to make it easier for researchers, students, and AI agents to understand and use a wide array of IR pipelines.

Figures (3)

• Figure 1: Example PyTerrier pipeline that performs retrieval, fusion, re-ranking, and answer generation stages. The pipeline is shown both as (a) declarative Python code and (b) the corresponding generated schematic.
• Figure 2: Example usage of the MCP server through the OpenAI API.
• Figure 3: A PyTerrier MCP server interaction example through Copilot.