search

Probable-Futures / climate-change-assistant

An experimental LLM agent to help people understand climate change. Not ready for use in a production environment.
0 stars 1 forks source link

readme

@probable-futures/climate-change-assistant

This project was created by Matthew Harris at DataKind using the Probable Futures open data resources. Matthew published a blog post about creating this project which contains extensive detail about his approach. This repository is a fork of Matthew's original repository and is not owned by Probable Futures, but rather is an experimental open source project that anyone is welcome to experiment with under Matthew's original terms.

[!CAUTION] This application is an experiment. Do not use it in publicly accessible applications. It should be used for experimental and research purposes only.

If you are looking for climate change data that is ready for public use, please see the Probable Futures climate handbook, climate maps, and open data resources. To learn more about Probable Futures, please visit the Probable Futures website.

About the app

This chainlit app will use an OpenAI Assistant and the Probable futures API to provide climate change information for a location, and provide some helpful resources for how to prepare.

The chainlit app is based off the cookbook example here.

screenshot

Setup

You will need a probable future API key, see here for more details. You will also need an OPENAI key.

Setup a Conda environment ...

  1. Install miniconda by selecting the installer that fits your OS version. Once it is installed you may have to restart your terminal (closing your terminal and opening again)
  2. cd app/
  3. conda env create -f environment.yml
  4. conda init zsh "zsh" can be replaced with your shell configuration
  5. conda activate assistant-env (You can leave the environment by using conda deactivate)

If you add a new dependency, run conda env update -f environment.yml to install it or to update existing ones.

Once you have these, then ...

  1. Copy .env.example to .env
  2. Set OpenAI key and Probable Futures API user and secret
  3. Create assistant with python3 create_assistant.py (If you have "ASSISTANT_ID" valid in .env you'll be updating that assistant)
  4. Copy the assistant id from output and put into .env

If you make changes to the assistant, rerun create_assistant.py, which will update the existing assistant. To create a new assistant, your "ASSISTANT_ID" in .env needs to be empty

To run chainlit app

chainlit run app.py

Run the app locally using docker (optional)

  1. Build the docker image docker build -t pf-assistant:latest .
  2. Run the app locally docker run -p 8080:8080 pf-assistant:latest

To view assistant on OpenAI

Go here