Forecats

Wake up every morning (more) excited to look at the forecast! This is a custom Home Assistant integration which uses forecast data, pictures of your pets, and Google's Nano Banana API to generate and serve weather-themed drawings of your precious babies every morning.

Blog post here.

Features

• Home assistant integration to generate pictures
• Customizable automation template to control drawing styles, generation times, etc.
• Optional optimization for E-ink screens (currently limited to spectra6 panels)

Examples

You can display it on an e-ink screen

Or on your HA Dashboard

Works in better weather

Getting Started

Warning

This will cost money. You get (I think) $300 of Gemini credits upon sign up, but eventually you will have to pay ~$0.14 every time this runs.

Requirements

• A home assistant server with the following add-ons
  • File editor add-on
  • Terminal and ssh add-on
  • Meteorologisk institutt (Met.no) integration
• A google AI studio API key. Note that, as of writing, you need to input billing details to get the free credits.

Setup

Do the following in your HA server, using the Terminal & SSH addon, or docker exec if you are running a container on a host system

1. Create the necessary directory structure in your Home Assistant server:

mkdir /config/custom_components && mkdir -p /config/forecats_data/input_images

2. Select and upload cat images

• Choose good pictures of your pets.
• Rename the files so the pets' names are in the filenames.
• Upload them to /config/forecats_data/input_images.

3. Download the repo

cd /config/custom_components && git clone https://github.com/jwardbond/forecats.git

4. Enable the custom integration by adding forecats: to your configuration file:

# configuration.yaml

default_config:

# ...existing data

automation: !include automations.yaml

forecats:

5. Set up the automation using the automation template (add to config/automations.yaml).

• Make sure to fill out (or remove) any <> in the template.

6. Restart your server

That's it! Every morning at 5:00 am, the forecats integration will generate the following images in the config/www/daily_forecats/ directory:

• forecats_original.png: the unprocessed output image from Gemini
• forecats_optimized.png: the output image cropped to your desired size and adjusted for display on your screen (currently only supports color adjustments for Spectra6 e-ink)

These images should be accessible on your local network at (e.g.): <YOUR HA URL>/local/daily_forecats/forecats_original.png

Note

It takes 10-30 seconds for gemini to generate the image. If you have any automations grabbing the image, then I recommend setting them to run a minute after the generate forecats autmation is set to run.

Note

To test the automation, go to developer tools > actions > generate forecats and run it manually.

Local Testing

I got annoyed testing out new prompts on HA, so I made a folder to experiment locally. If you would like to use it:

1. Clone the repo and enter the testing folder:

git clone https://github.com/jwardbond/forecats.git && cd forecats/local_testing

1. Add your cat images to the forecats_data/input_images folder.
2. Create a .env file with your Gemini API key.
3. Copy the data from your automation into test.py.
4. Run:

uv run test.py

(Optional) Sending to an e-ink screen

You will need a screen controllable with ESPHOME. I used seeed studio's e10002 spectra6 display. I've included the esphome config I use here. The basic idea is to set the automation to run every day at a 5:00 am, and have the screen wake up every day slightly before that, download the picture at 5:01 am (to leave time to generate), and then go into deep sleep until the next day.

TODO

• Enrol in HACS for easier install
• Option to save images to dir
• Make automation into blueprint for easier install
• Separate e-ink instructions
• See if I can make it more configurable from GUI
• Support for multiple cities