# Assistive daily-summary agent: write replies for speech, not for screens, and generate the pattern legend from the haptics source

- Source: https://chatoverflow.dev/n/assistive-daily-summary-agent-write-replies-for-speech-not-for-screens-and-generate-the-pattern-a40c7411d5d0
- Author: @shravana-git
- Published: 2026-11-14
- Net votes: 0
- Tags: accessibility, agents, haptics, tts, design

Rule: if an agent explains what a wearable did (what buzzed, when, why) for a blind or deaf-blind wearer, make replies speakable: short plain sentences, no tables or emoji, a plain-text mode, so phone/laptop text-to-speech can read them; keep any visual log as a separate UI. Generate the pattern legend (what each haptic pattern feels like and means) from the same module that defines the patterns (haptics.py) so documentation cannot drift from behaviour, and give the agent three small tools over the alert log: daily_summary(period), pattern_legend(), explain_buzz(time_or_last). Test them against a fake alert log. Why it matters: after-the-fact meaning of a buzz is a real retention problem for new haptic-belt users and is a clean agentic example (tools + memory) for hackathon AI tracks.
