preflight_voice.sh

#!/usr/bin/env bash
# preflight_voice.sh
# Quick health check for Christopher voice pipeline on Linux/WSL

set -u

WHISPER_BIN="${WHISPER_BIN:-$HOME/whisper.cpp/build/bin/whisper-cli}"
WHISPER_MODEL="${WHISPER_MODEL:-$HOME/whisper.cpp/models/ggml-base.en.bin}"
LLAMA_BIN="${LLAMA_BIN:-$HOME/llama.cpp/build/bin/llama-cli}"
PIPER_BIN="${PIPER_BIN:-piper}"
LISTEN_SECONDS="${LISTEN_SECONDS:-2}"

PASS_COUNT=0
WARN_COUNT=0
FAIL_COUNT=0

pass() { echo "[OK]   $1"; PASS_COUNT=$((PASS_COUNT + 1)); }
warn() { echo "[WARN] $1"; WARN_COUNT=$((WARN_COUNT + 1)); }
fail() { echo "[FAIL] $1"; FAIL_COUNT=$((FAIL_COUNT + 1)); }

section() { echo; echo "== $1 =="; }

detect_windows_host() {
  if [[ -n "${WINDOWS_HOST:-}" ]]; then
    echo "$WINDOWS_HOST"
    return
  fi

  if [[ -f /etc/resolv.conf ]]; then
    awk '/^nameserver / { print $2; exit }' /etc/resolv.conf
    return
  fi

  echo "127.0.0.1"
}

can_reach_pulse() {
  local host="$1"
  timeout 1 bash -c "cat < /dev/null > /dev/tcp/${host}/4713" >/dev/null 2>&1
}

choose_pulse_server() {
  local candidates=()
  local env_host=""

  if [[ -n "${PULSE_SERVER:-}" ]]; then
    env_host="${PULSE_SERVER#tcp:}"
    env_host="${env_host%%:*}"
    [[ -n "$env_host" ]] && candidates+=("$env_host")
  fi

  candidates+=("$(detect_windows_host)")

  if command -v ip >/dev/null 2>&1; then
    local gw
    gw="$(ip route 2>/dev/null | awk '/^default/ {print $3; exit}')"
    [[ -n "$gw" ]] && candidates+=("$gw")
  fi

  candidates+=("host.docker.internal" "127.0.0.1" "localhost")

  local seen=""
  local host
  for host in "${candidates[@]}"; do
    [[ -z "$host" ]] && continue
    case " $seen " in
      *" $host "*) continue ;;
    esac
    seen="$seen $host"

    if can_reach_pulse "$host"; then
      echo "tcp:${host}:4713"
      return
    fi
  done

  echo "tcp:$(detect_windows_host):4713"
}

cleanup() {
  [[ -n "${TMPDIR_AI:-}" && -d "$TMPDIR_AI" ]] && rm -rf "$TMPDIR_AI"
}
trap cleanup EXIT

TMPDIR_AI="$(mktemp -d)"
RAW_FILE="$TMPDIR_AI/input.raw"
WAV_FILE="$TMPDIR_AI/input.wav"
TRANSCRIPT_BASE="$TMPDIR_AI/transcript"

section "Environment"
WINDOWS_HOST="$(detect_windows_host)"
export PULSE_SERVER="$(choose_pulse_server)"
echo "PULSE_SERVER=$PULSE_SERVER"

section "Binaries + Models"
[[ -x "$WHISPER_BIN" ]] && pass "whisper-cli found: $WHISPER_BIN" || fail "whisper-cli missing or not executable: $WHISPER_BIN"
[[ -f "$WHISPER_MODEL" ]] && pass "Whisper model found: $WHISPER_MODEL" || fail "Whisper model missing: $WHISPER_MODEL"
[[ -x "$LLAMA_BIN" ]] && pass "llama-cli found: $LLAMA_BIN" || warn "llama-cli not found at expected path: $LLAMA_BIN"
command -v "$PIPER_BIN" >/dev/null 2>&1 && pass "piper found in PATH" || warn "piper not found in PATH"
command -v ffmpeg >/dev/null 2>&1 && pass "ffmpeg found" || warn "ffmpeg missing (needed for robust wav conversion)"

section "Audio Input"
CAPTURE_BACKEND="none"

if command -v parec >/dev/null 2>&1; then
  if timeout "$LISTEN_SECONDS" parec --format=s16le --rate=16000 --channels=1 > "$RAW_FILE" 2>/dev/null; then
    if [[ -s "$RAW_FILE" ]]; then
      CAPTURE_BACKEND="parec"
      pass "Mic capture works via PulseAudio (parec)"
    else
      warn "parec returned but captured empty audio"
    fi
  else
    warn "parec capture failed (PulseAudio may not be reachable)"
  fi
else
  warn "parec not installed"
fi

if [[ "$CAPTURE_BACKEND" == "none" ]] && command -v arecord >/dev/null 2>&1; then
  if arecord -f S16_LE -r 16000 -c 1 -d "$LISTEN_SECONDS" "$WAV_FILE" >/dev/null 2>&1; then
    if [[ -s "$WAV_FILE" ]]; then
      CAPTURE_BACKEND="arecord"
      pass "Mic capture works via ALSA (arecord)"
    else
      warn "arecord returned but captured empty audio"
    fi
  else
    warn "arecord capture failed"
  fi
fi

[[ "$CAPTURE_BACKEND" == "none" ]] && fail "No working microphone capture backend found"

section "ASR Smoke Test"
if [[ "$CAPTURE_BACKEND" == "parec" ]]; then
  if command -v ffmpeg >/dev/null 2>&1; then
    ffmpeg -y -f s16le -ar 16000 -ac 1 -i "$RAW_FILE" "$WAV_FILE" >/dev/null 2>&1 || true
  elif command -v sox >/dev/null 2>&1; then
    sox -t raw -r 16000 -e signed -b 16 -c 1 "$RAW_FILE" "$WAV_FILE" >/dev/null 2>&1 || true
  fi
fi

if [[ -s "$WAV_FILE" && -x "$WHISPER_BIN" && -f "$WHISPER_MODEL" ]]; then
  if "$WHISPER_BIN" -m "$WHISPER_MODEL" -f "$WAV_FILE" --output-txt --output-file "$TRANSCRIPT_BASE" --no-timestamps -t 4 >/dev/null 2>&1; then
    if [[ -f "${TRANSCRIPT_BASE}.txt" ]]; then
      pass "Whisper ASR executed successfully"
    else
      warn "Whisper ran but transcript file was not produced"
    fi
  else
    fail "Whisper ASR execution failed"
  fi
else
  warn "Skipping ASR smoke test (no wav captured or missing Whisper assets)"
fi

section "Audio Output"
if command -v paplay >/dev/null 2>&1; then
  if head -c 22050 /dev/zero | paplay --raw --format=s16le --rate=22050 --channels=1 >/dev/null 2>&1; then
    pass "Playback path works via paplay"
  else
    warn "paplay exists but playback test failed"
  fi
elif command -v ffplay >/dev/null 2>&1; then
  if ffplay -f lavfi -i anullsrc=r=22050:cl=mono -t 0.3 -nodisp -autoexit >/dev/null 2>&1; then
    pass "Playback path works via ffplay"
  else
    warn "ffplay exists but playback test failed"
  fi
else
  warn "No playback tool found (paplay or ffplay)"
fi

section "Summary"
echo "Pass: $PASS_COUNT | Warn: $WARN_COUNT | Fail: $FAIL_COUNT"

if [[ $FAIL_COUNT -gt 0 ]]; then
  echo
  echo "Fixes to try:"
  echo "1) Start Windows PulseAudio: C:\\PulseAudio\\start-pulseaudio.cmd"
  echo "2) Export PULSE_SERVER in WSL:"
  echo "   export PULSE_SERVER=tcp:$(awk '/^nameserver / {print $2; exit}' /etc/resolv.conf):4713"
  echo "3) Re-run: ./preflight_voice.sh"
  exit 1
fi

echo "Voice preflight passed. You can now run ./voice_ai.sh"
exit 0