NOTE: In branch "chatgpt_conversation_module", I've added a ChatGPT module that leverages the OpenAI GPT-4 model to enable the Garmin user to ask any questions while being on open seas.
Before setting out on our year-long sailing voyage, we needed a solution for accessing crucial weather data while being out on the open sea without internet connection. With this handy tool, we can now pull GRIB files on-demand using the Garmin inReach device, which turned out to be a cost-effective workaround to pricier options like IridiumGo. The system's entire process is illustrated in the following diagram:
Credit goes to Rhycus for the initial idea behind this workflow (https://github.com/rhycus/GRIB-via-inReach). Rather than merely sending specific location data and compressed wind details, this code goes a step further by transmitting the entire GRIB file. This provides access to a broader range of weather parameters, including pressure and waves. Additionally, the enhanced compression techniques allow for a similar number of messages and the weather data can now be analysed by using a GRIB viewer app that can interpolate data across different time points and zoom levels.
Feel free to use and modify the code to your needs. However, keep in mind that it depends on multiple APIs and libraries, which are subject to change. Always review the code thoroughly before using it, and proceed at your own risk!
- GARMIN INREACH: Make sure you have a Garmin inReach device with an active subscription. The unlimited message plan is recommended: https://www.garmin.com/en-US/p/837461/pn/010-06005-SU
- GMAIL SETUP: Create a dedicated Gmail account. Make sure the Gmail API is set up and you have the credentials.json files downloaded: https://developers.google.com/gmail/api/quickstart/python. Make sure to change the release status from test to production, otherwise authorisation will expire after 7 days!
- HOSTING: Use PythonAnywhere or another hosting service to run your script.
- DEVICE SETUP: Ensure you have a device like a tablet or computer with Jupyter Notebook to run the decoder and a GRIB file viewer app.
To request a GRIB-file, send a message to the dedicated Gmail adress via the Garmin inReach with specified weather model, location range, grid size, times and weather paramters. Here's an example of such a request:
ecmwf:24n,34n,72w,60w|8,8|12,48|wind,press
This translates to a request for data based on the ECMWF model, covering latitudes from 24N to 34N and longitudes from 72W to 60W, sampled at 8-degree intervals. The forecast times are 12 and 48 hours, and the requested weather parameters are wind and pressure.
The PythonAnywhere-based receiving service routinely checks the Gmail inbox at one-minute intervals for new requests. This task is managed by the continuous operation of the main.py file. Upon identifying a new message, it forwards the request to the Saildocs email API (http://www.saildocs.com/gribinfo). Saildocs promptly responds by sending an email with the requested GRIB file attached to our Gmail.
Subsequently, the binary content of the GRIB file is extracted, compressed (zipped), and encoded using base64. This compression step effectively reduces the file size by approximately 35%. The processed data is then divided into smaller chunks, prepared for transmission back to the inReach device. The transmission occurs via a post-request, utilising the designated inReach link provided with the initial request message.
NOTE: Base 64 encoding uses a character set of {A–Z, a–z, 0–9, +, /}, making it suitable for message transmission. While a base 85 representation could further compress the data, reducing its size by an additional 10%, it involves many special characters. These require extra attention. For instance, Rhycus faced issues with certain character combinations like '>f' that were unsendable and demanded extra handling through character shift which may result in sending more messages than anticipated.
NOTE: As Rhycus pointed out, messages occasionally get truncated around the 130-140 character range. To circumvent this, we've capped each message to 120 characters.
The messages received on the inReach device appear as follows (based on the request above):
Message 1
msg 0/1: <--- Indicates beginning of the first message
eJxzD/J0YmCIY2RgkGFKmvW/QTGTgUucQ5yBgZGBh4uBgUEUxGJQYPjPwMDCwMwQe6BR0qGBYY5Dw+4GeQd5BwcGMBBjYWA45Fx+jV3uZOtdi80Mdvu4FcyB
end <--- Indicates end of the first message
Message 2
msg 1/1: <--- Indicates beginning of the second message
wB236QYkmu62I4u94O5mA9bUhKCWPw0I0xPgpiuR5XYJDqDp0ZmvOAJaPKsjPDonyjAo7mEgYD4JroeYr1/FoZPsla2rF7Zisi3DUrD5AIClUBs=
end <--- Indicates end of the second message
To decode these messages efficiently, open them via the Garmin Earthmate App on an iPad. Then, copy and paste the messages into the Decoder Jupyter Notebook, for instance, using the Carnets App. Once decoded, you can view the resulting GRIB-file with a GRIB viewer app, such as LuckGrib.
Utilising this method, you can acquire wind and pressure data for the Atlantic crossing with two time points with just 6 messages.
Request:
ecmwf:44n,10n,75w,10w|8,8|12,48|wind,press
Result (in GRIB viewer app):
