AWI GPT

Four tools - ClimSight, pangaeaGPT, AWI_chatbot, and CMIP6 search - use LLMs to enhance research and operations. pangaeaGPT aids dataset exploration, ClimSight provides location-specific climate insights, AWI_chatbot streamlines internal tasks at AWI, and CMIP6 search improves access to CMIP6 data.

5
mentions
5
contributors

Cite this software

What AWI GPT can do for you

Our team developed four interconnected tools— ClimSight, pangaeaGPT, AWI_chatbot, and CMIP6 search—that use Large Language Models (LLMs) to tackle challenges in scientific research and organizational efficiency
These tools significantly benefit research and operations.

ClimSight is an open-source software tool designed to make localized climate information both accessible and actionable. It addresses the challenge of obtaining detailed local climate projections and understanding their impacts on specific activities. By integrating large language models (LLMs) with geographical and climate data, ClimSight provides users with tailored, location-specific insights essential for a variety of applications.

screencast

The primary function of ClimSight is to serve as a localized climate decision-making assistant. Users can access current and projected climate data for specific locations, enabling them to evaluate the potential effects of climate change on sectors such as agriculture, urban planning, and renewable energy. By delivering tailored and actionable insights, ClimSight generates customized summaries and recommendations based on user queries and location-specific data.

Methodologically, ClimSight orchestrates the retrieval and synthesis of data from multiple sources, including local climate conditions and future projections. It employs standard packages like LangGraph and LangChain for agent management and LLM integration, which promotes interoperability with other systems. The software processes user queries through the LLM, which interprets the request, fetches relevant climate data, and generates a comprehensive response that includes analyses and recommendations.

The architecture of ClimSight is modular, ensuring that components can be reused and adapted for various applications. This design choice enhances the software’s sustainability and relevance over time, allowing for easy updates and integration with new technologies or data sources. While currently utilizing OpenAI’s models, ClimSight is designed for flexibility; it can be transitioned to other open-source LLMs, ensuring adaptability and encouraging community-driven enhancements.

pangaeaGPT allows efficient exploration of PANGAEA datasets, enabling search, analysis, and visualization via advanced AI models. It employs a multi-agent architecture, including specialized agents for dataset retrieval, dataframe analysis, and visualization—both general-purpose and domain-specific agents. A supervisor agent coordinates these components to effectively address user queries, tackling challenges in navigating complex environmental data.

AWI_chatbot, for internal use at the Alfred Wegener Institute (AWI), assists employees with tasks by accessing internal documentation via Retrieval-Augmented Generation (RAG) technology. It streamlines operations with accurate responses to inquiries, such as setting up printers or arranging business trips.

CMIP6 search is specifically designed to improve access to CMIP6 data. It enhances the ability of researchers to search, request, and retrieve climate model data from the Coupled Model Intercomparison Project Phase 6 (CMIP6), streamlining access to critical climate datasets for research and analysis. By leveraging LLMs, this tool makes the process of obtaining and working with CMIP6 data faster and more efficient, reducing complexity for researchers.

Our software serves distinct groups: pangaeaGPT aids researchers in accessing and analyzing environmental datasets; ClimSight helps policymakers obtain reliable climate change information for local decisions; AWI_chatbot assists AWI staff by simplifying access to internal information; CMIP6 search aids researchers in efficiently accessing and utilizing CMIP6 climate model data for advanced climate research.

Participating organisations

Alfred Wegener Institute for Polar and Marine Research (AWI)

Reference papers

Mentions

Contributors

NK
Nikolay V. Koldunov
Alfred-Wegener-Institut für Polar und Meeresforschung
AJ
Antonia Anna Jost
Alfred-Wegener-Institut für Polar und Meeresforschung
BS
Boris Shapkin
Alfred-Wegener-Institut Helmholtz-Zentrum für Polar- und Meeresforschung
DP
Dmitrii Pantiukhin
Alfred-Wegener-Institut Helmholtz-Zentrum für Polar- und Meeresforschung
IK
Ivan Kuznetsov
Alfred Wegener Institute for Polar and Marine Research