ESS-DIVE AI Assistant
Connect ESS-DIVE datasets to AI-powered tools to retrieve, search, and interpret datasets.
What is the AI Assistant?
The AI Assistant is a tool that can help you search, download, and interpret ESS-DIVE data with the help of any AI model of your choosing. The assistant will help AI models properly fetch metadata, generate citations, look up reporting format data files on ESS-DIVE, and more through systems like Codex, Claude Code and VS Code.
Figure 1 below demonstrates what it would be like to have the assistant find ESS-DIVE datasets by data type and summarize the search results.
We also have written Skills that can be used independent of our assistant to help AI model perform specific tasks related to ESS-DIVE.
The program needs to be installed and connected to an AI Client, such as Goose, with an AI Agent, such as Gemini. The program allows you the flexibility to choose your preferred AI model to answer questions and perform tasks within your preferred AI client interface.
👉 To find the best starting point, check out our Setup section below. Otherwise, head over to ESS-DIVE's GitHub Repository to jump right in.
The most basic set up through Goose takes around 20 minutes.
What will you need?
🔑 Access to an AI Agent API key
💻 Client desktop app (e.g. Goose, VSCode)
🤖 Install AI Assistant
Agent vs. Agent Client Terminology
Agent: An AI model capable of determining and executing goals and steps to take based on prompts.
Agent Client: The interface where you interact with an AI agent. E.g. Goose, Codex, Claude Code.
Why use ESS-DIVE's AI Assistant?
The type of tasks needed for interacting with data on ESS-DIVE are not natively supported in large language model (LLM) chat interfaces and would require a lot of assistance to complete a task. Once set up, the assistant can:
Access data in consistent, predictable ways without guesswork
Provide context from a variety of sources
Work with variations
Interpret results
What tasks can the AI Assistant help you with?
Searching public ESS-DIVE datasets
Fetching dataset metadata, version history, and sharing permissions
Converting between ESS-DIVE dataset IDs and DOIs
Generating consistent ESS-DIVE data citations with MCP/API access details, plus warning-backed Crossref fallback for non-ESS-DIVE DOIs
Parsing File Level Metadata (FLMD) CSV content
Searching with Deep Dive API to find information within Reporting Format files
Looking up ESS-DIVE project acronyms, descriptions, and portal URLs
Turning coordinates into map links
Always make sure to cite your data! Datasets are valuable research contributions that should be cited to ensure that authors receive credit for producing and curating data. Always retain citations when using data from a published ESS-DIVE dataset.
The AI Assistant can help provide you with citations.
Example prompt you can ask the AI: Could you please dump the citations for these datasets to a
TSV? Use ESS-DIVE format.
How does the AI Assistant work?
The AI Assistant is a combination between an MCP (Model Context Protocol), an AI agent (e.g. Chat GPT, Gemini) and AI client of your choice (e.g. Claude, Codex, Goose). An MCP-Server is a software that provides instructions, via tools and skills, to AI agents telling it how to accomplish tasks.
The ESS-DIVE AI Assistant is an MCP-Server written by ESS-DIVE developers to instruct AI on the proper use of ESS-DIVE's APIs. When you install the AI Assistant, you are installing an MCP-Server.
The AI Assistant is not a web interface that you can interact with in your browser.
Tool
The MCP Server operates by having a collection of tools, or functions written in python, which tell the AI agent how to use ESS-DIVE's APIs and execute specific tasks. The tools are what enable the actions listed under "What can the AI Assistant Help you With?".
AI agents will know when to use ESS-DIVE's tools based on the context within your prompt. It won't be necessary to specify which tool you want it to use.
Skill
Skills provide natural language instructions for accomplishing common tasks and are written in markdown following an open standard understood by most agent clients.
Skills can be used with or without the MCP Server. Start off without Skills, then add them later if you want more predictable agent behavior for repeated tasks. Or the Skills can still be installed as reusable instruction documents outside of ESS-DIVE's MCP Server.
The MCP server provides the tools. A Skill helps the agent decide how and when to use them.
Setup
The GitHub Documentation provides multiple options and detailed instructions to set up your AI Assistant + Skills based on your preferences. Select which part of the setup documentation you would like to jump to:
Last updated