diff --git a/assets/alarm/alarm_test.lab b/assets/alarm/alarm_test.lab
new file mode 100644
index 0000000..cf732a8
--- /dev/null
+++ b/assets/alarm/alarm_test.lab
@@ -0,0 +1,55 @@
+0 700000 j
+700000 1440000 u
+1440000 2370000 u
+2370000 3110000 i
+3110000 3710000 ch
+3710000 4380000 i
+4380000 4800000 g
+4800000 5560000 a
+5560000 6180000 ts
+6180000 6850000 u
+6850000 7500000 j
+7500000 8210000 u
+8210000 9130000 u
+9130000 9880000 i
+9880000 10460000 ch
+10460000 11150000 i
+11150000 11690000 g
+11690000 12470000 a
+12470000 13100000 ts
+13100000 13770000 u
+13770000 14420000 j
+14420000 15140000 u
+15140000 16070000 u
+16070000 16810000 i
+16810000 17420000 ch
+17420000 18080000 i
+18080000 18610000 g
+18610000 19410000 a
+19410000 20020000 ts
+20020000 20680000 u
+20680000 21320000 j
+21320000 22030000 u
+22030000 22900000 u
+22900000 23640000 i
+23640000 24250000 ch
+24250000 24920000 i
+24920000 25460000 g
+25460000 26200000 a
+26200000 26840000 ts
+26840000 27480000 u
+27480000 28130000 j
+28130000 28830000 u
+28830000 29720000 u
+29720000 30440000 i
+30440000 31040000 ch
+31040000 31750000 i
+31750000 32600000 by
+32600000 33320000 o
+33320000 34120000 o
+34120000 34740000 j
+34740000 35350000 a
+35350000 35870000 s
+35870000 36510000 u
+36510000 36960000 t
+36960000 38220000 o
diff --git a/assets/alarm/alarm_test.wav b/assets/alarm/alarm_test.wav
new file mode 100644
index 0000000..24fc245
Binary files /dev/null and b/assets/alarm/alarm_test.wav differ
diff --git a/assets/img/idle.png b/assets/img/idle.png
new file mode 100644
index 0000000..f853294
Binary files /dev/null and b/assets/img/idle.png differ
diff --git a/assets/img/on_alarm.png b/assets/img/on_alarm.png
new file mode 100644
index 0000000..a7512fc
Binary files /dev/null and b/assets/img/on_alarm.png differ
diff --git a/audio_server.py b/audio_server.py
new file mode 100644
index 0000000..43265b6
--- /dev/null
+++ b/audio_server.py
@@ -0,0 +1,112 @@
+"""
+Alarm audio streaming test server.
+
+Streams a WAV file as raw PCM chunks over WebSocket on port 8766.
+Repeats every 30-60 seconds to exercise the ESP32 audio pipeline.
+
+Protocol:
+  1. Text frame:  {"type":"alarm_start","sample_rate":N,"channels":N,"bits":N}
+  2. Binary frames: raw PCM chunks (4096 bytes each, paced at ~90% real-time)
+  3. Text frame:  {"type":"alarm_stop"}
+"""
+
+import asyncio
+import json
+import logging
+import struct
+import wave
+from pathlib import Path
+from random import randint
+
+import websockets
+
+logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(message)s")
+log = logging.getLogger("audio_server")
+
+PORT = 8766
+CHUNK_SIZE = 4096
+AUDIO_DIR = Path(__file__).parent / "assets" / "alarm"
+
+
+def find_wav() -> Path:
+    """Find the first .wav file in the alarm assets directory."""
+    wavs = list(AUDIO_DIR.glob("*.wav"))
+    if not wavs:
+        raise FileNotFoundError(f"No .wav files found in {AUDIO_DIR}")
+    log.info("Using audio file: %s", wavs[0].name)
+    return wavs[0]
+
+
+def read_wav(path: Path) -> tuple[bytes, int, int, int]:
+    """Read WAV file and return (pcm_data, sample_rate, channels, bits_per_sample)."""
+    with wave.open(str(path), "rb") as wf:
+        sr = wf.getframerate()
+        ch = wf.getnchannels()
+        bits = wf.getsampwidth() * 8
+        pcm = wf.readframes(wf.getnframes())
+    log.info("WAV loaded: %dHz %dch %dbit, %.1fs, %d bytes",
+             sr, ch, bits, len(pcm) / (sr * ch * (bits // 8)), len(pcm))
+    return pcm, sr, ch, bits
+
+
+def chunk_bytes(data: bytes, size: int):
+    """Yield data in fixed-size chunks."""
+    for i in range(0, len(data), size):
+        yield data[i : i + size]
+
+
+async def stream_alarm(ws, pcm: bytes, sr: int, ch: int, bits: int):
+    """Stream one alarm cycle to the connected client."""
+    # Compute pacing: how long each chunk represents in seconds
+    bytes_per_sec = sr * ch * (bits // 8)
+    chunk_duration = CHUNK_SIZE / bytes_per_sec
+    pace_delay = chunk_duration * 0.9  # 90% real-time to avoid underrun
+
+    total_chunks = (len(pcm) + CHUNK_SIZE - 1) // CHUNK_SIZE
+
+    # Start
+    start_msg = json.dumps({
+        "type": "alarm_start",
+        "sample_rate": sr,
+        "channels": ch,
+        "bits": bits,
+    })
+    await ws.send(start_msg)
+    log.info("Sent alarm_start (%d chunks, pace %.1fms)", total_chunks, pace_delay * 1000)
+
+    # Stream PCM chunks
+    for i, chunk in enumerate(chunk_bytes(pcm, CHUNK_SIZE)):
+        await ws.send(chunk)  # bytes → binary frame
+        await asyncio.sleep(pace_delay)
+
+    # Stop
+    await ws.send(json.dumps({"type": "alarm_stop"}))
+    log.info("Sent alarm_stop")
+
+
+async def handler(ws):
+    """Handle a single WebSocket connection."""
+    remote = ws.remote_address
+    log.info("Client connected: %s:%d", remote[0], remote[1])
+
+    wav_path = find_wav()
+    pcm, sr, ch, bits = read_wav(wav_path)
+
+    try:
+        while True:
+            delay = randint(30, 60)
+            log.info("Next alarm in %ds", delay)
+            await asyncio.sleep(delay)
+            await stream_alarm(ws, pcm, sr, ch, bits)
+    except websockets.exceptions.ConnectionClosed:
+        log.info("Client disconnected: %s:%d", remote[0], remote[1])
+
+
+async def main():
+    log.info("Audio server starting on port %d", PORT)
+    async with websockets.serve(handler, "0.0.0.0", PORT):
+        await asyncio.Future()  # run forever
+
+
+if __name__ == "__main__":
+    main_loop = asyncio.run(main())
diff --git a/run_all.py b/run_all.py
new file mode 100644
index 0000000..a6ba7e9
--- /dev/null
+++ b/run_all.py
@@ -0,0 +1,15 @@
+#!/usr/bin/env python3
+"""Launch stats_server and audio_server as child processes."""
+import subprocess, sys, signal
+from pathlib import Path
+
+d = Path(__file__).parent
+procs = [
+    subprocess.Popen([sys.executable, d / "stats_server.py"]),
+    subprocess.Popen([sys.executable, d / "audio_server.py"]),
+]
+signal.signal(signal.SIGINT, lambda *_: [p.terminate() for p in procs])
+signal.signal(signal.SIGTERM, lambda *_: [p.terminate() for p in procs])
+print(f"Running stats_server (PID {procs[0].pid}) + audio_server (PID {procs[1].pid})")
+for p in procs:
+    p.wait()