Wander: A Screen-Free Voice Companion That Gets You Outside 🌿

A developer built Wander, a screen-free, voice-first walking companion that runs its reasoning loop locally using Gemma 3 via Ollama and faster-whisper, with ElevenLabs or an offline Piper fallback for voice replies. The FastAPI server narrates nearby places and species, issues "touch grass" challenges, maintains a life list of sightings, plans walking loops, and logs walks to a map journal, using a POI cache so walks can continue with almost no internet after the first fetch. The project was submitted to the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass.

Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass Submission 🌿 This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass https://dev.to/challenges/hacktoberfest-week1-2026-10-05 Wander is a screen-free, voice-first walking companion. You put your phone in your pocket, tap once, and talk. It narrates nearby places and species, gives you small “touch grass” challenges, keeps a life list of what you spot, plans walking loops, and logs every walk to a map journal. The entire brain runs locally with Gemma 3 via Ollama + faster-whisper. Only the voice replies use ElevenLabs with an offline Piper fallback . Phone client ──WebSocket──▶ FastAPI server │ │ GPS + voice Whisper → WalkAgent → Gemma │ │ plays audio ◀── mp3 ──────── ElevenLabs / Piper Key pieces: poi cache – after the first fetch, the walk can continue with almost no internet. A closed cloud model would force every GPS ping and every utterance into someone else’s servers. Wander keeps the entire reasoning loop local so you can: The open pieces Gemma + local Whisper + open map/species data are what make the “phone stays in the pocket” promise actually work. This project is submitted under: Built during the challenge window 5–11 Oct 2026 . Thanks for reading — now go touch some grass.