Adhan Audio Caster

The Vision

Five salah times per day. Each one requires:

1. ▪Playing an audio Adhan on the TV at precisely the right time
2. ▪Pausing whatever media is playing on the Sony Android TV
3. ▪Showing a visual display on a Google Nest Hub Max
4. ▪Resuming media after the Adhan completes
5. ▪Doing all of this automatically, every day, on a $35 device

This is not a weekend project. This is systems engineering.

The Hardware Stack

• ▪Raspberry Pi 4 — the brain. Runs the Node.js orchestration server.
• ▪Sony Android TV — receives ADB commands over LAN for media control
• ▪Google Nest Hub Max — displays prayer-specific visuals via Cast API
• ▪Google Chromecast Audio — handles the audio casting pipeline

The Software Challenge

The hardest part wasn't any single piece — it was the coordination between them. ADB is finicky. The Nest Hub has no native API. Cast timing is unreliable. Media state detection is non-trivial.

My original estimate: 3 weeks to get a stable version.

Bringing in AI

I started working with Antigravity (Gemini 3 Flash/Pro) as an intelligent pair programmer. The difference was immediate.

Instead of spending 2 days researching ADB command sequences, I described the problem. The AI produced working ADB command chains with error handling in under 10 minutes. I spent my time testing and refining edge cases, not searching Stack Overflow.

# AI-generated ADB media pause/resume sequence
adb -s $TV_IP shell input keyevent KEYCODE_MEDIA_PAUSE
sleep $ADHAN_DURATION
adb -s $TV_IP shell input keyevent KEYCODE_MEDIA_PLAY

The Nest Hub dashboard was the biggest win. What I estimated as a 1-week design sprint became a 2-day build with the AI generating the CSS animations, prayer card layouts, and the dynamic dashboard refresh logic.

The 4-Day Sprint

• ▪Day 1: Architecture + ADB integration (AI-assisted)
• ▪Day 2: Nest Hub dashboard design (AI-generated CSS/animations)
• ▪Day 3: Prayer time calculation + scheduling (AI pair programming)
• ▪Day 4: Integration testing + production hardening (me, mostly)

Result: A production-grade IoT system running 24/7 for several months without failure.

What I Learned

AI doesn't replace engineering judgment. It accelerates the parts that are research-heavy and repetitive, freeing you to focus on the judgment calls that actually matter: system design, edge case handling, and knowing when something is "good enough for production."

---

Explore the source code on GitHub or see the system flow animation in the README.