I Built WanderWise: A Local-First AI Planner That Wants You to Close the Screen

Description and use

WanderWise uses a local MiniLM model to quickly and privately recommend short outdoor activities, cutting down screen time.

Problem

People find it hard to turn the advice “go outside” into a concrete, actionable plan, leading to more screen time and indecision.

Outcome

The app delivers a single, actionable outdoor plan that users can copy and immediately act on, reducing screen time and encouraging real-world activity.

Workflow

• 1. User enters intent, time, company, weather, energy. 2. Browser loads MiniLM embeddings via Transformers.js. 3. Intent and activity descriptions are embedded. 4. Cosine similarity scores activities. 5. Hybrid ranking adds constraint scores (duration, weather, energy, company). 6. Top activity is presented with three timed steps, gear list, safety cue, and copy button. 7. User copies plan and closes the tab.

Evidence

WanderWise AI turns a sentence such as “I need a calm reset after work and I have one hour” into a small, practical outdoor plan.

Source

The model runs in the browser. It converts the user's natural-language intention and each activity description into normalized 384-dimensional embeddings.

Source

This split gives the model the job it is good at—understanding intent—while ordinary code handles time, conditions, and safety.

Source