GrassWhisper: A Local AI Gatekeeper That Lets You Touch Grass 🌿

Description and use

GrassWhisper locally classifies walk‑time notifications, speaks only the important ones, aggregates bursts, and provides a walk summary, reducing screen dependence.

Problem

Users receive notifications while walking but must look at their phones to decide which ones matter, increasing screen time and distraction.

Outcome

GrassWhisper reduces screen dependence by filtering and speaking only critical or important notifications during walks, aggregates bursts, respects privacy, and provides a summary of the walk, thereby encouraging users to stay outside.

Workflow

• Activate Walk Mode (manual or motion‑sensor detection)
• Receive incoming notifications
• Send notification data to local AI (Ollama) with Walk Mode context
• Apply deterministic safety and priority rules to AI output
• Assign urgency level (Critical, Important, Queued)
• Aggregate similar notifications
• Speak alerts via Web Speech API (respecting Privacy Mode)
• Log walk duration, screen time, notifications received, allowed, and suppressed
• Generate post‑walk No Screen Challenge summary

Evidence

GrassWhisper classifies notifications into three levels: 🔴 Critical — speak immediately, 🟠 Important — give a brief interruption, 🟢 Queued — stay silent and save it for later

Source

It also goes beyond one‑notification‑at‑a‑time triage.

Source

GrassWhisper is built around local open‑source AI inference rather than sending notification content to a cloud AI service.

Source

The notification pipeline is therefore: Notification → Local AI + rules → Urgency level → Aggregation → Privacy‑safe audio → Post‑walk digest

Source