Cartesi Weather AI weather prediction

Project built for the Cartesi Hackathon Inteli 2024 edition

CLI fork instalation

First of all, you must install the Foundry SDK:

curl -L https://foundry.paradigm.xyz | bash

It's also important you have the docker daemon running. If you don't, open up Docker Desktop.

In order to run our project, you must use our fork of the Cartesi CLI (or else the container won't run) before proceeding to run the dapp inside the Cartesi Virtual Machine.

1. Use NVM or some virtual environment to avoid conflicts with Cartesi's CLI package. Example:

   nvm install 20.14
   nvm use 20.14
   

   Make sure that running cartesi yields an error saying command not found.
2. Clone the repo sarmentow/cli
3. Install the pnpm package manager with

   curl -fsSL https://get.pnpm.io/install.sh | sh -
   

   (Any instalation problems with pnpm, refer to their docs
4. Inside the CLI directory run:

   pnpm install
   pnpm run build
   

5. Then, symlink the resulting executables to /usr/local/bin so that you can run it from anywhere:

   sudo ln -s $(pwd)/apps/cli/bin/run.js /usr/local/bin/cartesi
   

If you now type cartesi help you should see the cli output text in your terminal

Installation

To setup data encryption, first generate a private key for the Cartesi Machine:

openssl genrsa -out cartesi-weather/dapp/private_key.pem 4096

You also need to generate a private key for the client so that it can decrypt the output from the Cartesi Machine model:

openssl genrsa -out cartesi-weather/client/private_key.pem 4096

Then install the packages required for the client and server applications:

pip install numpy==1.26.4 Flask==3.0.3 gql==3.5.0 aiohttp web3 requests gradio folium tabulate scipy

You can configure the url for the server application through the SERVER_URL environment variable (default address is the same as the Flask development server default 127.0.0.1:5000)

Development setup

For better performance and faster iteration cycles, we strongly advise you to use Nonodo. In order to use it, simply run nonodo from any terminal session, and then, in a separate terminal, run python3 dapp.py from inside the cartesi-weather/dapp directory.

With both of these running, proceed to follow the Flask and Example instructions of the next section (ignore the part that uses the Cartesi CLI)

Usage with the Cartesi CLI

To run the program in your machine, run the following commands in separate terminal sessions, inside the cartesi-weather directory:

1. Start the API server

   flask --app service.py run
   

2. To run in production mode, go inside the cartesi-weather/dapp/ directory, and build the Cartesi rollup container image:

   cartesi build
   
   

   Then, run it

   cartesi run
   

3. Once you have your Cartesi rollup container running, you can run inference through the run_verifiable_inference and query_model_outputs methods. For a pre-defined example you can do

   python3 example.py
   

You should see a message saying that the application is running. Open http://localhost:7860/ in a web browser.

More info

Technical analysis/devlog