This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass
What I Built
I built a voice-assisted schedule reminder application that helps users stay organized and follow through on their planned activities. It keeps track of scheduled tasks and reminds users when it is time to act by playing voice reminders. The application also listens for verbal confirmation to determine whether the user has started the scheduled activity.
The goal is to help users spend less time glued to their screens and more time taking action in the real world. By combining voice reminders with audio-based confirmation, the app encourages users to stay on track, manage their time effectively, and focus on the tasks they planned ahead of time.
Demo
This is a link where you can get the mobile application from.
NOTE: the application works on mobile devices only
and all LLM downloads occur upon first installation and opening of the application so ensure you have a minimum of 500MB of data
Code
Toluwaloju0
/
mlh_touch_grass
This repository is for the mlh project touch grass. i have tried creating an executable which users can run on their local machine or devices
How I Built It
I built the application using Python 3.12 in a virtual environment, with the Codex coding agent assisting me throughout the development process.
For the voice features, I integrated Kokoro ONNX for text-to-speech, allowing the application to deliver scheduled reminders through voice playback. I also used FLAN-T5-Small as part of the language-processing pipeline for handling user confirmations.
I developed the mobile application's user interface with Kivy, a Python framework for building cross-platform applications, and used Buildozer to package the project for Android.
My goal was to create a lightweight, voice-assisted application that helps users manage their schedules, respond to reminders, and follow through on their planned activities without constantly interacting with their screens.
Why Does Open Innovation Matter?
Open innovation makes it possible to bring AI-powered voice reminders to Android devices without relying on constant internet access or expensive cloud APIs. My project aims to help users stay on schedule, spend less time on screens, and take action on their planned activities while making AI more accessible to people with limited connectivity.
My Agent Session
I want to build a mobile app using Python and Kivy. Create a voice reminder app based on schedules saved by the user. Support one schedule for all days or individual schedules per weekday via a toggle. Store schedules in JSON. Include Home navigation, a schedule editor, and a Usage page. At schedule time, speak the activity and listen for one minute for agreement. Use Kokoro ONNX for speech output and an offline speech-to-text model; keep processing offline after first-run model downloads.
Implemented a Kivy app in main.py with Home, Schedule, and Usage pages; daily or weekday-specific schedules saved to local JSON; first-run downloads for Kokoro ONNX and Vosk; and local yes/no phrase classification. FLAN-T5-small is not an ASR model, so Vosk was chosen for speech recognition. The in-process checker is only best-effort: reliable Android background reminders need a foreground service, alarm and reboot handling, permissions, and Android-compatible audio/ONNX dependencies.
Edit buildozer.spec so this application can be packaged as a mobile .apk.
Updated buildozer.spec for the Voice Schedule app, Android INTERNET and RECORD_AUDIO permissions, API 35/minimum API 24, ARM architectures, source exclusions, and APK outputs for debug and release. Kokoro ONNX, Vosk, and sounddevice were not added as requirements because their Android compiled dependencies need compatible builds or python-for-android recipes. A debug build can be started with: buildozer -v android debug. No APK build was run.
How do I create a DevRelay agent session for this chat?
Explained that the user can ask Codex to save the chat as a DevRelay agent session, or submit a normalized transcript using the DevRelay CLI.
Save this chat as a DevRelay agent session. Follow this DEV submission format: What I Built (what it does, how it gets people off-screen and outdoors, who it is for); Demo; Code; How I Built It (open-source AI and local inference); Why Open Innovation Matters; My Agent Session; Prize Categories. Note: no demo link or repository was provided.
Top comments (0)