TrailBuddy: less scrolling, more fresh air
I had a pretty simple idea for this challenge: if I have a free hour and want to get outside, why does deciding what to do sometimes take longer than the walk? So I made TrailBuddy. You put in a starting area, how much time you have, and what you feel like doing. It gives you three ideas, a rough duration, and one thing to bring. The aim is to pick one and close the tab — not spend the afternoon…
A simple concept inspired the creation of TrailBuddy: if an individual possesses an hour to venture outdoors, why should selecting an activity demand more time than the actual walk? Consequently, the TrailBuddy application was developed. Users input a starting location, the available time, and preferred activity. In response, the AI generates three suggestions, an approximate duration, and an essential item to carry.
The objective is for users to select one suggestion and promptly close the tab, rather than dedicating an entire afternoon to planning. The AI operates locally on the user's personal device instead of utilizing a centralized AI service. By employing Qwen 2.5 3B through Ollama on the user's own machine, the user can avoid the need for an API key or per-request model charges, and maintain their location and interests on their own device.
The prompt for the AI is available in the public repository, and users have the option to switch to a different compatible Ollama model by setting an environment variable. However, a crucial prerequisite exists: the necessary model and Python packages must first be downloaded. Once these are in place, the application can function without an internet connection.
If the Ollama service is not running, TrailBuddy defaults to a selection of pre-generated suggestions, which are labeled as "offline" and are not AI-generated. The application was constructed using FastAPI, alongside plain HTML, CSS, and JavaScript. Currently, it does not incorporate features like map integration, weather updates, or real-time trail conditions, meaning the suggestions provided are merely ideas and not verified local recommendations.
Users are advised to verify the conditions before embarking on their journey. The application has undergone seven tests, with the local Ollama model not being available during the testing phase. The focus was placed on verifying the offline fallback feature. Additionally, the user has not yet tested the app in a real-world scenario.
They plan to do so and will update this report if they have the opportunity. The source code for TrailBuddy can be accessed at https://github.com/prarthanas9917/trailbuddy-ai, and this project was part of the Hacktoberfest Week 1 “Touch Grass” challenge.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
