Let's walk: An AI powered Scenic walking Companion

Quincy Aku-Boye — Sun, 11 Oct 2026 03:50:16 +0000

For the Hacktoberfest Open-Source AI Challenge, I vibe coded Let’s Walk with help from ChatGPT and GitHub Copilot (AI assistant), a walking companion that helps people plan a pedestrian loop around their available time, preferred scenery, and mood.

The idea is simple: a walk can be more than a line on a map. It can be an intentional exploration of your neighborhood, shaped by real route data, honest environmental information, and thoughtful AI reflection prompts.

Most mapping apps will happily draw you a route even if it is weak, vague, or poorly matched to what you actually want. Let’s Walk takes the opposite approach.

It uses real pedestrian routing through Valhalla, checks nearby scenery using OpenStreetMap, and explains the route with a local Gemma model running through Ollama. If a route isn’t ideal, or if the data isn’t available, the app says so instead of pretending everything is fine.

That honesty is a big part of the project for me. I wanted to build something that feels more respectful of the real world: a route should be based on actual walking constraints, not just a neat-looking line on a map.

How it works:

Pick your available walk time: 15, 30, or 45 minutes.
Choose the kind of scenery you want: greenery, parks, gardens, water, quiet streets, and more.
The app finds a real walking loop using Valhalla.
It checks nearby map features to understand what might be around the route.
A local AI model explains the route and generates reflection prompts for the walk.
You can then use the walking mode with a timer and spoken prompts to stay present while you move.
I built this to explore something I care about: how open-source AI can support calmer, more intentional experiences instead of only optimization and speed.

Built with:

JavaScript, CSS, and HTML on the frontend
Node.js on the backend
Valhalla for pedestrian routing
OpenStreetMap and Overpass for nearby scenery context
Ollama and Gemma 2 for local route explanations and prompts
A few design principles shaped the app:

Honesty over invention: if no good route is found, it says so. No fake routes. No pretending a park exists when it doesn’t.

Privacy-first thinking: the app runs locally, and the AI model is handled on the user’s machine rather than depending on a cloud service.

Real constraints matter: walking time, route geometry, and environmental clues are all treated as practical constraints instead of decorative extras.

Graceful degradation: if Ollama is unavailable or a data source is rate-limited, the app explains what happened and continues with safer fallback content.

I also wanted the app to feel a little more human. A walk is not just travel; it’s timing, mood, awareness, and attention. Let’s Walk tries to support that mood instead of turning the experience into a purely functional tool.

This is still a local MVP, and there are plenty of limitations. Public services like Valhalla and Overpass can be rate-limited or unavailable, and the app is built as a prototype for experimentation rather than production deployment. But that uncertainty is part of the point: the project treats missing data and limited services as a normal part of real-world software, not a bug to hide.

I’m excited by the possibility of building more thoughtful tools with open-source AI. Walking is one of the simplest and most human things we do, and I think it deserves software that respects that complexity.

If you try it out, I’d love feedback on:

the walking experience
the honesty-first design
what would make the app more useful in everyday life
whether local AI feels like an improvement here or just a gimmick
You can find the project here: https://github.com/akuquincy-commits/Lets-Walk

DEV Community: Quincy Aku-Boye

Let's walk: An AI powered Scenic walking Companion