TODO.md

@@ -9,30 +9,26 @@
 
 ### P1 · homeai-infra
 
-- [ ] Install Docker Desktop for Mac, enable launch at login
+- [x] Install Docker Desktop for Mac, enable launch at login
-- [ ] Create shared `homeai` Docker network
+- [x] Create shared `homeai` Docker network
-- [ ] Create `~/server/docker/` directory structure
+- [x] Create `~/server/docker/` directory structure
-- [ ] Write compose files: Home Assistant, Portainer, Uptime Kuma, Gitea, code-server, n8n
+- [x] Write compose files: Uptime Kuma, code-server, n8n (HA, Portainer, Gitea are pre-existing on 10.0.0.199)
-- [ ] Write `.env.secrets.example` and `Makefile`
+- [x] `docker compose up -d` — bring all services up
-- [ ] `make up-all` — bring all services up
+- [x] Home Assistant onboarding — long-lived access token generated, stored in `.env`
-- [ ] Home Assistant onboarding — generate long-lived access token
-- [ ] Write `~/server/.env.services` with all service URLs
 - [ ] Install Tailscale, verify all services reachable on Tailnet
-- [ ] Gitea: create admin account, initialise all 8 sub-project repos, configure SSH
+- [ ] Gitea: initialise all 8 sub-project repos, configure SSH
 - [ ] Uptime Kuma: add monitors for all services, configure mobile alerts
 - [ ] Verify all containers survive a cold reboot
 
 ### P2 · homeai-llm
 
-- [ ] Install Ollama natively via brew
+- [x] Install Ollama natively via brew
-- [ ] Write and load launchd plist (`com.ollama.ollama.plist`)
+- [x] Write and load launchd plist (`com.homeai.ollama.plist`) — `/opt/homebrew/bin/ollama`
-- [ ] Write `ollama-models.txt` with model manifest
+- [x] Register local GGUF models via Modelfiles (no download): llama3.3:70b, qwen3:32b, codestral:22b
-- [ ] Run `scripts/pull-models.sh` — pull all models
+- [x] Deploy Open WebUI via Docker compose (port 3030)
+- [x] Verify Open WebUI connected to Ollama, all models available
 - [ ] Run `scripts/benchmark.sh` — record results in `benchmark-results.md`
-- [ ] Deploy Open WebUI via Docker compose (port 3030)
-- [ ] Verify Open WebUI connected to Ollama, all models available
 - [ ] Add Ollama + Open WebUI to Uptime Kuma monitors
-- [ ] Add `OLLAMA_URL` and `OPEN_WEBUI_URL` to `.env.services`
 
 ---
 
@@ -40,20 +36,19 @@
 
 ### P3 · homeai-voice
 
-- [ ] Compile Whisper.cpp with Metal support
+- [x] Install `wyoming-faster-whisper` — model: faster-whisper-large-v3 (auto-downloaded)
-- [ ] Download Whisper models (`large-v3`, `medium.en`) to `~/models/whisper/`
+- [x] Install Kokoro ONNX TTS — models at `~/models/kokoro/`
-- [ ] Install `wyoming-faster-whisper`, test STT from audio file
+- [x] Write Wyoming-Kokoro adapter server (`homeai-voice/tts/wyoming_kokoro_server.py`)
-- [ ] Install Kokoro TTS, test output to audio file
+- [x] Write + load launchd plists for Wyoming STT (10300) and TTS (10301)
-- [ ] Install Wyoming-Kokoro adapter, verify Wyoming protocol
+- [x] Install openWakeWord + pyaudio — model: hey_jarvis
-- [ ] Write + load launchd plists for Wyoming STT (10300) and TTS (10301)
+- [x] Write + load openWakeWord launchd plist (`com.homeai.wakeword`)
-- [ ] Connect Home Assistant Wyoming integration (STT + TTS)
+- [x] Write `wyoming/test-pipeline.sh` — smoke test (3/3 passing)
+- [~] Connect Home Assistant Wyoming integration (STT + TTS) — awaiting HA UI config
 - [ ] Create HA Voice Assistant pipeline
 - [ ] Test HA Assist via browser: type query → hear spoken response
-- [ ] Install openWakeWord, test wake detection with USB mic
-- [ ] Write + load openWakeWord launchd plist
 - [ ] Install Chatterbox TTS (MPS build), test with sample `.wav`
 - [ ] Install Qwen3-TTS via MLX (fallback)
-- [ ] Write `wyoming/test-pipeline.sh` — end-to-end smoke test
+- [ ] Train custom wake word using character name
 - [ ] Add Wyoming STT/TTS to Uptime Kuma monitors
 
 ---

homeai-llm/launchd/com.homeai.ollama.plist

@@ -8,7 +8,7 @@
 
   <key>ProgramArguments</key>
   <array>
-    <string>/usr/local/bin/ollama</string>
+    <string>/opt/homebrew/bin/ollama</string>
     <string>serve</string>
   </array>
 

homeai-llm/modelfiles/Codestral-22B

@@ -0,0 +1,7 @@
+FROM /Users/aodhan/Models/LLM/Codestral-22B-v0.1-GGUF/Codestral-22B-v0.1-Q4_K_M.gguf
+
+PARAMETER num_ctx 16384
+PARAMETER temperature 0.2
+PARAMETER top_p 0.95
+
+SYSTEM "You are an expert coding assistant."

homeai-llm/modelfiles/Llama-3.3-70B

@@ -0,0 +1,7 @@
+FROM /Users/aodhan/Models/LLM/Llama-3.3-70B-Instruct-GGUF/Llama-3.3-70B-Instruct-Q4_K_M.gguf
+
+PARAMETER num_ctx 8192
+PARAMETER temperature 0.7
+PARAMETER top_p 0.9
+
+SYSTEM "You are a helpful AI assistant."

homeai-llm/modelfiles/Qwen3-32B

@@ -0,0 +1,7 @@
+FROM /Users/aodhan/Models/LLM/Qwen3-32B-GGUF/Qwen3-32B-Q4_K_M.gguf
+
+PARAMETER num_ctx 8192
+PARAMETER temperature 0.7
+PARAMETER top_p 0.9
+
+SYSTEM "You are a helpful AI assistant."

homeai-voice/scripts/launchd/com.homeai.wakeword.plist

@@ -0,0 +1,34 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN"
+  "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+  <key>Label</key>
+  <string>com.homeai.wakeword</string>
+
+  <key>ProgramArguments</key>
+  <array>
+    <string>/Users/aodhan/homeai-voice-env/bin/python3</string>
+    <string>/Users/aodhan/gitea/homeai/homeai-voice/wyoming/wakeword_daemon.py</string>
+    <string>--wake-word</string>
+    <string>hey_jarvis</string>
+    <string>--notify-url</string>
+    <string>http://localhost:8080/wake</string>
+  </array>
+
+  <key>RunAtLoad</key>
+  <true/>
+
+  <key>KeepAlive</key>
+  <true/>
+
+  <key>StandardOutPath</key>
+  <string>/tmp/homeai-wakeword.log</string>
+
+  <key>StandardErrorPath</key>
+  <string>/tmp/homeai-wakeword-error.log</string>
+
+  <key>ThrottleInterval</key>
+  <integer>10</integer>
+</dict>
+</plist>

homeai-voice/scripts/launchd/com.homeai.wyoming-stt.plist

@@ -0,0 +1,43 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN"
+  "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+  <key>Label</key>
+  <string>com.homeai.wyoming-stt</string>
+
+  <key>ProgramArguments</key>
+  <array>
+    <string>/Users/aodhan/homeai-voice-env/bin/wyoming-faster-whisper</string>
+    <string>--uri</string>
+    <string>tcp://0.0.0.0:10300</string>
+    <string>--model</string>
+    <string>large-v3</string>
+    <string>--language</string>
+    <string>en</string>
+    <string>--device</string>
+    <string>cpu</string>
+    <string>--compute-type</string>
+    <string>int8</string>
+    <string>--data-dir</string>
+    <string>/Users/aodhan/models/whisper</string>
+    <string>--download-dir</string>
+    <string>/Users/aodhan/models/whisper</string>
+  </array>
+
+  <key>RunAtLoad</key>
+  <true/>
+
+  <key>KeepAlive</key>
+  <true/>
+
+  <key>StandardOutPath</key>
+  <string>/tmp/homeai-wyoming-stt.log</string>
+
+  <key>StandardErrorPath</key>
+  <string>/tmp/homeai-wyoming-stt-error.log</string>
+
+  <key>ThrottleInterval</key>
+  <integer>10</integer>
+</dict>
+</plist>

homeai-voice/scripts/launchd/com.homeai.wyoming-tts.plist

@@ -0,0 +1,36 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN"
+  "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
+<plist version="1.0">
+<dict>
+  <key>Label</key>
+  <string>com.homeai.wyoming-tts</string>
+
+  <key>ProgramArguments</key>
+  <array>
+    <string>/Users/aodhan/homeai-voice-env/bin/python3</string>
+    <string>/Users/aodhan/gitea/homeai/homeai-voice/tts/wyoming_kokoro_server.py</string>
+    <string>--uri</string>
+    <string>tcp://0.0.0.0:10301</string>
+    <string>--voice</string>
+    <string>af_heart</string>
+    <string>--speed</string>
+    <string>1.0</string>
+  </array>
+
+  <key>RunAtLoad</key>
+  <true/>
+
+  <key>KeepAlive</key>
+  <true/>
+
+  <key>StandardOutPath</key>
+  <string>/tmp/homeai-wyoming-tts.log</string>
+
+  <key>StandardErrorPath</key>
+  <string>/tmp/homeai-wyoming-tts-error.log</string>
+
+  <key>ThrottleInterval</key>
+  <integer>10</integer>
+</dict>
+</plist>

homeai-voice/scripts/load-all-launchd.sh

@@ -0,0 +1,35 @@
+#!/usr/bin/env bash
+# Load (or reload) all homeai-voice launchd services.
+
+set -euo pipefail
+
+LAUNCHD_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/launchd" && pwd)"
+LAUNCH_AGENTS=~/Library/LaunchAgents
+
+PLISTS=(
+    com.homeai.wyoming-stt.plist
+    com.homeai.wyoming-tts.plist
+    com.homeai.wakeword.plist
+)
+
+for plist in "${PLISTS[@]}"; do
+    src="${LAUNCHD_DIR}/${plist}"
+    dst="${LAUNCH_AGENTS}/${plist}"
+    label="${plist%.plist}"
+
+    cp "$src" "$dst"
+
+    if launchctl list "$label" &>/dev/null; then
+        launchctl unload "$dst" 2>/dev/null || true
+    fi
+    launchctl load "$dst"
+    echo "Loaded: $label"
+done
+
+echo ""
+echo "Status:"
+for plist in "${PLISTS[@]}"; do
+    label="${plist%.plist}"
+    pid=$(launchctl list "$label" 2>/dev/null | awk 'NR==2{print $1}')
+    echo "  $label — PID: ${pid:--}"
+done

homeai-voice/tts/wyoming_kokoro_server.py

@@ -0,0 +1,145 @@
+#!/usr/bin/env python3
+"""Wyoming TTS server backed by Kokoro ONNX.
+
+Usage:
+    python wyoming_kokoro_server.py --uri tcp://0.0.0.0:10301 --voice af_heart
+"""
+
+import argparse
+import asyncio
+import logging
+import os
+
+import numpy as np
+
+from wyoming.audio import AudioChunk, AudioStart, AudioStop
+from wyoming.event import Event
+from wyoming.info import Attribution, Info, TtsProgram, TtsVoice, TtsVoiceSpeaker
+from wyoming.server import AsyncEventHandler, AsyncServer
+from wyoming.tts import Synthesize
+
+_LOGGER = logging.getLogger(__name__)
+
+SAMPLE_RATE = 24000
+SAMPLE_WIDTH = 2  # int16
+CHANNELS = 1
+CHUNK_SECONDS = 1  # stream in 1-second chunks
+
+
+def _load_kokoro():
+    from kokoro_onnx import Kokoro
+    model_dir = os.path.expanduser("~/models/kokoro")
+    return Kokoro(
+        os.path.join(model_dir, "kokoro-v1.0.onnx"),
+        os.path.join(model_dir, "voices-v1.0.bin"),
+    )
+
+
+class KokoroEventHandler(AsyncEventHandler):
+    def __init__(self, tts, default_voice: str, speed: float, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._tts = tts
+        self._default_voice = default_voice
+        self._speed = speed
+
+        # Send info immediately on connect
+        asyncio.ensure_future(self._send_info())
+
+    async def _send_info(self):
+        info = Info(
+            tts=[
+                TtsProgram(
+                    name="kokoro",
+                    description="Kokoro ONNX TTS",
+                    attribution=Attribution(
+                        name="thewh1teagle/kokoro-onnx",
+                        url="https://github.com/thewh1teagle/kokoro-onnx",
+                    ),
+                    installed=True,
+                    voices=[
+                        TtsVoice(
+                            name=self._default_voice,
+                            description="Kokoro voice",
+                            attribution=Attribution(name="kokoro", url=""),
+                            installed=True,
+                            languages=["en-us"],
+                            speakers=[TtsVoiceSpeaker(name=self._default_voice)],
+                        )
+                    ],
+                )
+            ]
+        )
+        await self.write_event(info.event())
+
+    async def handle_event(self, event: Event) -> bool:
+        if Synthesize.is_type(event.type):
+            synthesize = Synthesize.from_event(event)
+            text = synthesize.text
+            voice = self._default_voice
+
+            if synthesize.voice and synthesize.voice.name:
+                voice = synthesize.voice.name
+
+            _LOGGER.debug("Synthesizing %r with voice=%s speed=%.1f", text, voice, self._speed)
+
+            try:
+                loop = asyncio.get_event_loop()
+                samples, sample_rate = await loop.run_in_executor(
+                    None, lambda: self._tts.create(text, voice=voice, speed=self._speed)
+                )
+
+                samples_int16 = (np.clip(samples, -1.0, 1.0) * 32767).astype(np.int16)
+                audio_bytes = samples_int16.tobytes()
+
+                await self.write_event(
+                    AudioStart(rate=SAMPLE_RATE, width=SAMPLE_WIDTH, channels=CHANNELS).event()
+                )
+
+                chunk_size = SAMPLE_RATE * SAMPLE_WIDTH * CHANNELS * CHUNK_SECONDS
+                for i in range(0, len(audio_bytes), chunk_size):
+                    await self.write_event(
+                        AudioChunk(
+                            rate=SAMPLE_RATE,
+                            width=SAMPLE_WIDTH,
+                            channels=CHANNELS,
+                            audio=audio_bytes[i : i + chunk_size],
+                        ).event()
+                    )
+
+                await self.write_event(AudioStop().event())
+                _LOGGER.info("Synthesized %.1fs of audio", len(samples) / sample_rate)
+
+            except Exception:
+                _LOGGER.exception("Synthesis error")
+                await self.write_event(AudioStop().event())
+
+        return True  # keep connection open
+
+
+async def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--uri", default="tcp://0.0.0.0:10301")
+    parser.add_argument("--voice", default="af_heart")
+    parser.add_argument("--speed", type=float, default=1.0)
+    parser.add_argument("--debug", action="store_true")
+    args = parser.parse_args()
+
+    logging.basicConfig(
+        level=logging.DEBUG if args.debug else logging.INFO,
+        format="%(asctime)s %(levelname)s %(name)s %(message)s",
+    )
+
+    _LOGGER.info("Loading Kokoro ONNX model...")
+    tts = _load_kokoro()
+    _LOGGER.info("Kokoro loaded. Starting Wyoming TTS on %s (voice=%s)", args.uri, args.voice)
+
+    server = AsyncServer.from_uri(args.uri)
+
+    def handler_factory(reader, writer):
+        return KokoroEventHandler(tts, args.voice, args.speed, reader, writer)
+
+    await server.run(handler_factory)
+
+
+if __name__ == "__main__":
+    asyncio.run(main())

homeai-voice/wyoming/test-pipeline.sh

@@ -0,0 +1,88 @@
+#!/usr/bin/env bash
+# Smoke test for the Wyoming voice pipeline.
+# Tests: STT server alive | TTS server alive | TTS generates audio
+#
+# Usage: ./test-pipeline.sh
+
+set -euo pipefail
+
+STT_HOST="${STT_HOST:-localhost}"
+STT_PORT="${STT_PORT:-10300}"
+TTS_HOST="${TTS_HOST:-localhost}"
+TTS_PORT="${TTS_PORT:-10301}"
+VENV="${VENV:-$HOME/homeai-voice-env}"
+
+PASS=0
+FAIL=0
+
+check() {
+    local name="$1"; local result="$2"
+    if [[ "$result" == ok* ]]; then
+        echo "  [PASS] $name${result#ok}"; PASS=$((PASS + 1))
+    else
+        echo "  [FAIL] $name — $result"; FAIL=$((FAIL + 1))
+    fi
+}
+
+echo "=== HomeAI Voice Pipeline Smoke Test ==="
+echo ""
+
+echo "1. STT Wyoming server (port $STT_PORT)"
+if nc -z -w2 "$STT_HOST" "$STT_PORT" 2>/dev/null; then
+    check "STT port open" "ok"
+else
+    check "STT port open" "port $STT_PORT not reachable — is wyoming-stt running?"
+fi
+
+echo ""
+echo "2. TTS Wyoming server (port $TTS_PORT)"
+if nc -z -w2 "$TTS_HOST" "$TTS_PORT" 2>/dev/null; then
+    check "TTS port open" "ok"
+else
+    check "TTS port open" "port $TTS_PORT not reachable — is wyoming-tts running?"
+fi
+
+echo ""
+echo "3. Kokoro TTS synthesis test"
+TTS_OUTPUT="/tmp/homeai-tts-test.wav"
+"$VENV/bin/python3" - <<'PYEOF'
+import sys, os, asyncio
+import numpy as np
+
+model_dir = os.path.expanduser("~/models/kokoro")
+model_path = os.path.join(model_dir, "kokoro-v1.0.onnx")
+voices_path = os.path.join(model_dir, "voices-v1.0.bin")
+
+if not os.path.exists(model_path):
+    print(f"Model not found: {model_path}")
+    sys.exit(1)
+
+from kokoro_onnx import Kokoro
+tts = Kokoro(model_path, voices_path)
+samples, sr = tts.create("Hello, I am your home assistant. The voice pipeline is working.", voice="af_heart", speed=1.0)
+
+# Write WAV
+import wave, struct
+samples_int16 = (np.clip(samples, -1.0, 1.0) * 32767).astype(np.int16)
+with wave.open("/tmp/homeai-tts-test.wav", "w") as wf:
+    wf.setnchannels(1)
+    wf.setsampwidth(2)
+    wf.setframerate(sr)
+    wf.writeframes(samples_int16.tobytes())
+
+print(f"Generated {len(samples)/sr:.1f}s of audio at {sr}Hz")
+sys.exit(0)
+PYEOF
+if [[ $? -eq 0 && -f "$TTS_OUTPUT" ]]; then
+    size=$(wc -c < "$TTS_OUTPUT")
+    check "Kokoro synthesis" "ok — ${size} bytes written to $TTS_OUTPUT"
+    echo ""
+    echo "  To play: afplay $TTS_OUTPUT"
+else
+    check "Kokoro synthesis" "synthesis failed"
+fi
+
+echo ""
+echo "─────────────────────────────────"
+echo "Results: $PASS passed, $FAIL failed"
+[[ $FAIL -eq 0 ]]

homeai-voice/wyoming/wakeword_daemon.py

@@ -0,0 +1,110 @@
+#!/usr/bin/env python3
+"""Always-on wake word detection daemon using openWakeWord.
+
+Listens on the default microphone, fires an HTTP POST to --notify-url
+when the wake word is detected.
+
+Usage:
+    python wakeword_daemon.py --wake-word hey_jarvis --notify-url http://localhost:8080/wake
+"""
+
+import argparse
+import logging
+import time
+import urllib.request
+import json
+import numpy as np
+
+_LOGGER = logging.getLogger(__name__)
+
+SAMPLE_RATE = 16000
+CHUNK_SIZE = 1280  # ~80ms at 16kHz — recommended by openWakeWord
+
+
+def notify(url: str, wake_word: str, score: float):
+    payload = json.dumps({"wake_word": wake_word, "score": float(score)}).encode()
+    try:
+        req = urllib.request.Request(
+            url,
+            data=payload,
+            headers={"Content-Type": "application/json"},
+            method="POST",
+        )
+        with urllib.request.urlopen(req, timeout=2):
+            pass
+        _LOGGER.info("Wake word '%s' detected (score=%.3f) — notified %s", wake_word, score, url)
+    except Exception as e:
+        _LOGGER.warning("Failed to notify %s: %s", url, e)
+
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--wake-word", default="hey_jarvis")
+    parser.add_argument("--notify-url", default="http://localhost:8080/wake")
+    parser.add_argument("--threshold", type=float, default=0.5)
+    parser.add_argument("--cooldown", type=float, default=3.0, help="Seconds between triggers")
+    parser.add_argument("--model-dir", default=None, help="Path to custom .onnx wake word model")
+    parser.add_argument("--debug", action="store_true")
+    args = parser.parse_args()
+
+    logging.basicConfig(
+        level=logging.DEBUG if args.debug else logging.INFO,
+        format="%(asctime)s %(levelname)s %(message)s",
+    )
+
+    try:
+        import pyaudio
+    except ImportError:
+        _LOGGER.error("pyaudio not installed. Run: pip install pyaudio")
+        raise SystemExit(1)
+
+    import openwakeword
+    from openwakeword.model import Model
+
+    _LOGGER.info("Loading wake word model: %s", args.wake_word)
+
+    model_paths = []
+    if args.model_dir:
+        import os, glob
+        model_paths = glob.glob(os.path.join(args.model_dir, "*.onnx"))
+
+    oww = Model(
+        wakeword_models=model_paths if model_paths else [args.wake_word],
+        inference_framework="onnx",
+    )
+
+    audio = pyaudio.PyAudio()
+    stream = audio.open(
+        rate=SAMPLE_RATE,
+        channels=1,
+        format=pyaudio.paInt16,
+        input=True,
+        frames_per_buffer=CHUNK_SIZE,
+    )
+
+    _LOGGER.info("Listening for wake word '%s'...", args.wake_word)
+    last_trigger = 0.0
+
+    try:
+        while True:
+            raw = stream.read(CHUNK_SIZE, exception_on_overflow=False)
+            chunk = np.frombuffer(raw, dtype=np.int16)
+            oww.predict(chunk)
+
+            for ww, scores in oww.prediction_buffer.items():
+                score = scores[-1] if scores else 0.0
+                if score >= args.threshold:
+                    now = time.time()
+                    if now - last_trigger >= args.cooldown:
+                        last_trigger = now
+                        notify(args.notify_url, ww, score)
+    except KeyboardInterrupt:
+        pass
+    finally:
+        stream.stop_stream()
+        stream.close()
+        audio.terminate()
+
+
+if __name__ == "__main__":
+    main()