foundational-nature-ai

Explore the current LLMs as a prototype for a Foundational Nature AI capable of informing questions about biodiversity and conservation relevant for real-world decisions.

---

Setup

The environment

To run the notebooks you need to create an environment with the dependencies. There are two options:

Docker

If you have docker in your system, you run a jupyter lab server with:

docker compose up --build

And if you want to get into the container, use a terminal in jupyter lab, vscode remote development or run this command:

docker exec -it foundational_nature_ai_notebooks /bin/bash

Conda environment

Create the environment with:

mamba env create -n foundational_nature_ai -f environment.yml

This will create an environment called foundational-nature-ai with a common set of dependencies.

API Keys

You will need to generate appropriate API keys for the LLMs that you want to use. Create a .env file and save this in the root directory, e.g. OPENAI_API_KEY={YOUR-KEY}

Data

Download the data, eval and training folders from here: https://drive.google.com/drive/folders/1idjUhOEqePqDYp-l-jUFq2jl7v-8yCm5?usp=sharing

and place in the root of the project

Model and evalaution setup

Parameters for defining which model to run the evaluation/app with are defined in the params dictionary constructed in setup.py

Streamlit app

To run the steamlit app execute:

streamlit run src/app.py

This will run a prototype app and launch it in a web browser