# Community call: OpenRefine LLM Extension Demo

**URL:** <https://forum.openrefine.org/t/community-call-openrefine-llm-extension-demo/2923>\
**Category:** Events\
**Tags:** community-calls\
**Created:** [June 18, 2026, 2:29pm UTC](https://forum.openrefine.org/t/community-call-openrefine-llm-extension-demo/2923 "2026-06-18T14:29:01Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Martin](https://dub1.discourse-cdn.com/flex017/user_avatar/forum.openrefine.org/martin/32/7_2.png) [@Martin](https://forum.openrefine.org/u/Martin)\
**Post date:** [June 18, 2026, 2:29pm UTC](https://forum.openrefine.org/t/community-call-openrefine-llm-extension-demo/2923/1 "2026-06-18T14:29:01Z")

</div>

We are organizing an OpenRefine community call focused on concrete uses of the [OpenRefine LLM Extension](https://github.com/sunilnatraj/llm-extension).

This session is intended for anyone interested in using LLMs with OpenRefine. The goal is to show practical examples of how the extension can support data-wrangling workflows, including use cases from data journalism and library metadata work.

- **Date and time:** 2026-07-07T12:00:00Z → 2026-07-07T13:30:00Z
- **Facilitator:** @Martin
- **Presenters:** @h_piedcoq @psm @Sunil_Natraj
- **[Optional registration](https://5no9et.share-na3.hsforms.com/2qCE6OV2fRGupN361Cu2i3g)** to receive a reminder
- **[Zoom link](https://us02web.zoom.us/j/82539835762)**

This session will be recorded and shared on YouTube.

## Session Overview

The [OpenRefine LLM Extension](https://github.com/sunilnatraj/llm-extension) makes it possible to call large language models in OpenRefine workflows. The session will show two examples of how LLMs are used in journalism and GLAM use cases, including:

- Preparing and transforming data for journalistic work
- Exploring library-specific metadata workflows
- Using local or external providers, including Ollama and other LLM services
- Identifying what works well and what still needs improvement

Questions from users, trainers, extension developers, and contributors are welcome.

## Agenda

- Short introduction of each participant
- Community announcements
- Demo by @h_piedcoq based on his [DataHarvest 2026](https://forum.openrefine.org/t/tutorial-openrefine-and-llm-extension/2883) tutorial, with a focus on data journalism workflows
- Demo by @psm on the use of OpenRefine with the AI plugin for library-specific purposes, including integrations with Ollama and other providers
- Questions and discussion with @Sunil_Natraj, developer of the OpenRefine LLM Extension

---

<div class="post-metadata">

**Author:** ![Martin](https://dub1.discourse-cdn.com/flex017/user_avatar/forum.openrefine.org/martin/32/7_2.png) [@Martin](https://forum.openrefine.org/u/Martin)\
**Post date:** [July 27, 2026, 6:10pm UTC](https://forum.openrefine.org/t/community-call-openrefine-llm-extension-demo/2923/2 "2026-07-27T18:10:55Z")

</div>

The recording of the OpenRefine Community Call: **OpenRefine LLM Extension Demo** is now available on YouTube:

https://www.youtube.com/embed/UnAv9DDVAVI?feature=oembed&wmode=opaque

Thank you again to @h_piedcoq, @psm, and @Sunil_Natraj for contributing to the session.

## Recording overview

The recording includes three main parts:

- **Hervé Piedcoq’s demo** (starts at 00:19 min), based on his [DataHarvest 2026 tutorial](https://github.com/herve-checkfirst/DataHarvest2026-Refine_with_llm), with a focus on data journalism workflows. Hervé spent a significant part of the demo showing name cleaning with the LLM extension. His tutorial materials are linked in the first post of this thread.

- **Parthasarathi Mukhopadhyay’s demo** (starts at 24:23 min), focused on library-specific use cases and testing different models via Ollama. His related materials are available here:

- **Community discussion** (starts at 50:15 min), including questions with Sunil Natraj, developer of the OpenRefine LLM Extension.

## Topics covered

Some of the points discussed during the session include:

- using the LLM extension for name cleaning and data journalism workflows
- testing local models through Ollama
- library metadata use cases
- using structured JSON responses so that an LLM can return multiple data points in a single answer
- including fields such as confidence scores or reasoning notes in structured outputs
- possible uses of LLMs to review reconciliation service candidates