CropWizard uses the power of Generative AI to support data-driven decision making for agriculture
The CropWizard Advisor is a natural language chat interface that harnesses the power of proprietary and open Large Language Models (LLMs) to provide answers to technical questions about US and world agriculture, including questions with text, images, and other input data.
CropWizard uses Retrieval Augmented Generation (RAG) to consult a large database of over 650,000 Extension publications, research papers, and other technical documents when responding to user questions. It provides references to the original source documents used in its responses to enable users to explore the topic in more detail and to verify details in the responses against trusted sources.
CropWizard can also consult a variety of computational tools such as simple calculators, advanced neural network models, and data agents, to answer questions that pertain to those tools. Tools and agents within CropWizard are used to obtain weather information for a location, analyze shapefiles for a farm, recognize pests in images using a pretrained computer vision model, analyze data in spreadsheets, and search for public information on the Web, among other features. Ongoing research in the CropWizard team is adding new multimodal question-answering capabilities and training new foundation models for data-driven decision making using crop models and reinforcement learning.
CropWizard is built on the remarkable Illinois Chat framework — the easiest way we know of to build interactive AI systems using RAG, tools, reasoning, and multimodal analysis. The software is open source and available on Github at CropWizard and Illinois Chat, including extensive documentation. The software and documentation are free for research, teaching, and other non-commercial uses, but commercial use requires a paid license. Contact digitalag@illinois.edu if you are interested in licensing CropWizard.