Why I built TrailBuddy Local

Modern apps optimize for attention, but I wanted an outdoor companion that helps people step away from glowing screens instead of pulling them back in. TrailBuddy Local is a local-first AI companion for mindful walks: it creates a personalized mission, keeps the experience on-device, and never requires cloud AI or GPS tracking.

What the project does

  • Generates short outdoor missions from time budget, interest, difficulty, and mobility constraints
  • Uses a local Ollama model to produce safety-aware prompts and reflection tasks
  • Keeps a private SQLite journal for completed missions and reflection notes
  • Enforces loopback-only endpoints and blocks cloud inference by design

Why open-source AI matters

This project uses an open-weight model and local inference so the app works offline and keeps every decision and personal reflection on the user's machine. For outdoor guidance, privacy matters as much as usefulness: a walk should not require uploading personal habits or location data to a third-party provider. Open-source AI makes that possible without locking the experience behind a paid API or cloud dependency.

How it was built

TrailBuddy Local is a Streamlit app that runs entirely on the user's machine. The project pairs a local Ollama model with schema validation, safety guardrails, and a SQLite-backed local journal. The privacy layer rejects non-loopback endpoints and keeps the app working in airplane mode after setup.

TrailBuddy Local: Touch Grass Challenge Session
You

Submit TrailBuddy Local to the Hacktoberfest Open-Source AI Challenge: Week 1, save this agent session, and create a DEV post draft.

Agent

I reviewed the project and confirmed it fits the challenge: a privacy-first AI outdoor companion that runs local models only, keeps data on-device, and helps users spend less time on screens.

▸ repo_inspection README.md + app.py + src/*
Input
README.md + app.py + src/*
Output
TrailBuddy Local is a Streamlit app that creates mindful outdoor missions using local Ollama inference and a local SQLite journal.
Agent

I installed the Python dependencies and ran the project validation. The remaining setup gate is the local Ollama runtime and model, which this challenge expects to be available on the user's machine.

▸ python_tests python -m pytest -q
Input
python -m pytest -q
Output
The app setup is valid but local integration tests are blocked until Ollama is running and qwen2.5:3b is available.
3 of 3 messages

Project links

This submission is a fresh, local-first Hacktoberfest project built around the challenge prompt: open-source AI at its center, with practical value for real people in the real world.