With GLM 5.2 being released recently, I decided to give it a try with an idea I had. I use a closed-source workout tracking app called Lyfta (https://lyfta.app). However it recently occurred to me that I have no access or control over my data on their platform, if their servers went down then I'd lose my past few months of progress. So, I did the next best thing, I reverse engineered their APK, sniffed live network requests via Frida and chucked it all into a neat Caddy + ExpressJS wrapper. Now I have the ability to save my workout information and pictures, and also forward them to a Telegram channel. I had to completely redevelop the 'muscle group highlights' on the body model completely itself, as that's handled client-side and not by Lyfta's server, but after MUCH digging through decompiled Smali files from the APK, I got it nailed down fairly close to how it originally looked. This project was mostly done by GLM 5.2 and I am incredibly impressed at how it's been handling my tasks so far, phenomenal performance compared to when I had access to Fable. This project also probably isn't authorized by Lyfta's API terms of service but whatever, sue me ¯\_(ツ)_/¯
kivir•1h ago
With GLM 5.2 being released recently, I decided to give it a try with an idea I had. I use a closed-source workout tracking app called Lyfta (https://lyfta.app). However it recently occurred to me that I have no access or control over my data on their platform, if their servers went down then I'd lose my past few months of progress. So, I did the next best thing, I reverse engineered their APK, sniffed live network requests via Frida and chucked it all into a neat Caddy + ExpressJS wrapper. Now I have the ability to save my workout information and pictures, and also forward them to a Telegram channel. I had to completely redevelop the 'muscle group highlights' on the body model completely itself, as that's handled client-side and not by Lyfta's server, but after MUCH digging through decompiled Smali files from the APK, I got it nailed down fairly close to how it originally looked. This project was mostly done by GLM 5.2 and I am incredibly impressed at how it's been handling my tasks so far, phenomenal performance compared to when I had access to Fable. This project also probably isn't authorized by Lyfta's API terms of service but whatever, sue me ¯\_(ツ)_/¯
This project is also deployed on the Docker Hub too at: https://hub.docker.com/repository/docker/wolfgirl/openlyfta