Extension AI

“Extension AI” work is supported by funding from a cooperative agreement with USDA-NIFA and in partnership with the University of New Hampshire, New Technologies for Ag Extension, grant no. 2023-41595-41325 . This work includes the ExtensionBot tool that the foundation is working on in the Artificial Intelligence space. The Large Language and Embedding models used by ExtensionBot can also be utilized in other AI-based tools and services that we hope to develop and deploy in the coming months and years.

What is ExtensionBot?

ExtensionBot is a LLM Agnostic chatbot that is built and deployed by the Extension Foundation. We are working closely with Eduworks Corporation on the development and deployment of the technology. ExtensionBot is backed by an LLM (Large Language Model) that is trained exclusively on data provided by Cooperative Extension institutions and Ask Extension. Because ExtensionBot provides citations to Extension resources along with its responses, these responses can be verified in real-time. Our platform is built so that LLMs can quickly be retrained and replaced depending on performance and application.

Our larger goal is to build an LLM that is trained on Extension Resources from across the Cooperative Extension service. Each new institution or contributing member strengthens the LLM. In addition to asking institutions to create data dumps, we are asking they create a “data pipeline” for their data. This is an API endpoint that is updated on a regular basis so our LLM is consistently being updated.

Listen to a great use case of how Oklahoma State is using ExtensionBot.

What is MERLIN?

MERLIN (Machine-driven Extension Research and Learning Innovation Network Platform) is a data management platform powering ExtensionBot and other AI-driven initiatives within the Extension Foundation. MERLIN provides essential tools for data organization, validation, and accessibility, ensuring a seamless AI-driven knowledge delivery system. Key features include:

  • Managing data crawling and API endpoints
  • Allowing users to view and manage crawled datasets and endpoint data
  • Handling data validation and integrity checks
  • Displaying the last crawl date and last ExtensionBot ingestion date
  • Providing access to ExtensionBot chat logs

Look for more information on MERLIN in spring of 2025

Try it out!

Experiment with ExtensionBot

ExtensionBot

Contributing institutions along with the number of resources they have provided.

Oklahoma State University (1,540)
Oregon State University (5,203)
University of Georgia (1100)
University of Florida Publications (6,402)
University of Florida Blogs(10,354)
University of Tennessee(2,602)
LSU (1917)
Penn St. (4,740) In development
University of California – Integrated Pest Management (1,675)
University of Nevada at Reno (1232)
North Dakota State University (1535)
Ohio State (1262)
sare.org
Ask Extension Knowledgebase (313,443)

University of Maryland (331)
Colorado State University (602)
Mississippi State University (4,215)
Clemson University (997)
Alabama A&M + Auburn (2,772)
Utah State University (1,197)
University of Arkansas (1,175)
North Carolina State University (1,900)
Virginia Tech (300)
Tennessee State University (180)
University of Arizona (2028)
Wisconsin (1257)
Illinois(3137)
West Virginia(3280)
University of Kentucky(2610)
University of Michigan(9699)

Extension Foundation

AI Advisory Board

This board will consist of members of the Extension Foundation and participating institutions. This board will help guide policy, feature development and implementation of AI tools such as ExtenstionBot. Additionally, this group can help guide the future of AI within the broader Cooperative Extension Service.

Interested in Contributing to ExtensionBot?

Follow these steps:

  • STEP

    1

    DECIDE

    Decide what content you would like to include/contribute to ExtensionBot. This can include documents like publications and factsheets.

  • STEP

    2

    CREATE

    Do you have the ability to create a data dump?

    YES – Ask your local IT/Web development staff to create a dump based on our json specification. Once it passes validation, then share it with Mark Locklear for processing.

    NO – The Extension Foundation may be able to crawl your Extension website. Contact Mark Locklear to chat about the location and format of your data.

  • STEP

    3

    INTERNAL TESTING

    Configure ExtBot locally and do some internal testing. You may need to tweak your configuration based on the results. Visit chat.extension.org for implantation details and options.

  • STEP

    4

    ADD EXTBOT TO YOUR WEBSITE

    Once you are ready to go live you can implement either a full page version of ExtBot, the widget, or both. Reporting is available for you to view the interactions your users are having with ExtBot.

Customization and Localization

When you implement ExtensionBot on your website you can modify the colors and logo to match your institution.

Additionally, you can configure how ExtensionBot responds to questions so that it favors your institutional dataset.

For example, a user asks a question about switchgrass. If you have content or a resource on this topic, then your institutional content can be favored and ExtensionBots response will be based on your local content in addition to being cited in the response. However, if no local content is found that is relevant to the users question, then the system can respond to the users inquiry based on other content within the system. This is all 100% configurable on your end.

For more information contact Mark Locklear or David Warren.

Sign Up for our Newsletter to Receive Updates About ExtensionBot!

This is not a marketing newsletter, but an opportunity for the Extension Foundation to provide you with updates including release notes, bug fixes, and upcoming features.