Merge pull request #11 from legop3/laptoprover

Laptoprover

pi/bin/audio-forward-listener.sh

@@ -2,20 +2,70 @@
 set -euo pipefail
 set +H
 
-ENV_FILE="${VIDEO_ENV_FILE:-/var/lib/roverd/video.env}"
+ENV_FILE="${ROVERD_MEDIA_ENV_FILE:-/var/lib/roverd/media.env}"
 
 if [[ ! -f "$ENV_FILE" ]]; then
   echo "Environment file ${ENV_FILE} missing; cannot start audio forward listener" >&2
   exit 1
 fi
 
-# shellcheck disable=SC1090
-source "$ENV_FILE"
+# Load KEY=VALUE pairs from media.env without evaluating shell syntax. The
+# forward URL is data produced by roverd, and treating it as shell code would
+# break on normal SRT query-string characters such as '&'.
+load_env_file() {
+  local content=""
 
-: "${AUDIO_FORWARD_URL:?AUDIO_FORWARD_URL not set in ${ENV_FILE}}"
-PLAYBACK_DEVICE="${AUDIO_PLAYBACK_DEVICE:-forward}"
-AUDIO_NORMALIZE_ENABLE="${AUDIO_NORMALIZE_ENABLE:-1}"
-AUDIO_NORMALIZE_FILTER="${AUDIO_NORMALIZE_FILTER:-dynaudnorm=f=75:g=15:m=10:p=0.9,alimiter=limit=0.85:level=disabled}"
+  if [[ -r "$ENV_FILE" ]]; then
+    content="$(cat "$ENV_FILE")"
+  elif command -v sudo >/dev/null 2>&1; then
+    content="$(sudo -n cat "$ENV_FILE" 2>/dev/null || true)"
+  fi
+
+  if [[ -z "$content" ]]; then
+    echo "Cannot read ${ENV_FILE} (permission denied). Run as a user that can read it, or allow sudo -n for cat." >&2
+    exit 1
+  fi
+
+  local line key val
+  while IFS= read -r line || [[ -n "$line" ]]; do
+    [[ "$line" =~ ^[[:space:]]*$ ]] && continue
+    [[ "$line" =~ ^[[:space:]]*# ]] && continue
+
+    if [[ "$line" =~ ^[[:space:]]*export[[:space:]]+([A-Za-z_][A-Za-z0-9_]*)=(.*)$ ]]; then
+      key="${BASH_REMATCH[1]}"
+      val="${BASH_REMATCH[2]}"
+    elif [[ "$line" =~ ^[[:space:]]*([A-Za-z_][A-Za-z0-9_]*)=(.*)$ ]]; then
+      key="${BASH_REMATCH[1]}"
+      val="${BASH_REMATCH[2]}"
+    else
+      continue
+    fi
+
+    val="${val#${val%%[![:space:]]*}}"
+    val="${val%${val##*[![:space:]]}}"
+    if [[ "$val" =~ ^\".*\"$ ]]; then
+      val="${val:1:${#val}-2}"
+    elif [[ "$val" =~ ^\'.*\'$ ]]; then
+      val="${val:1:${#val}-2}"
+    fi
+
+    printf -v "$key" '%s' "$val"
+    export "$key"
+  done <<< "$content"
+}
+
+load_env_file
+
+: "${ROVERD_AUDIO_PLAYBACK_ENABLE:?ROVERD_AUDIO_PLAYBACK_ENABLE not set in ${ENV_FILE}}"
+if [[ "${ROVERD_AUDIO_PLAYBACK_ENABLE}" -ne 1 ]]; then
+  echo "Audio playback disabled by roverd media config; skipping audio forward listener" >&2
+  exit 0
+fi
+: "${ROVERD_AUDIO_PLAYBACK_FORWARD_URL:?ROVERD_AUDIO_PLAYBACK_FORWARD_URL not set in ${ENV_FILE}}"
+: "${ROVERD_AUDIO_PLAYBACK_DEVICE:?ROVERD_AUDIO_PLAYBACK_DEVICE not set in ${ENV_FILE}}"
+: "${ROVERD_AUDIO_PLAYBACK_NORMALIZE:?ROVERD_AUDIO_PLAYBACK_NORMALIZE not set in ${ENV_FILE}}"
+: "${ROVERD_AUDIO_PLAYBACK_NORMALIZE_FILTER:?ROVERD_AUDIO_PLAYBACK_NORMALIZE_FILTER not set in ${ENV_FILE}}"
+PLAYBACK_DEVICE="${ROVERD_AUDIO_PLAYBACK_DEVICE}"
 
 if [[ -n "${FFMPEG_BIN:-}" ]]; then
   FFMPEG_BIN_PATH="$FFMPEG_BIN"
@@ -45,12 +95,12 @@ run_pipeline() {
     -flags low_delay
     -analyzeduration 200k
     -probesize 32k
-    -i "${AUDIO_FORWARD_URL}"
+    -i "${ROVERD_AUDIO_PLAYBACK_FORWARD_URL}"
     -vn
   )
 
-  if [[ "${AUDIO_NORMALIZE_ENABLE}" -ne 0 ]]; then
-    ffmpeg_args+=(-af "${AUDIO_NORMALIZE_FILTER}")
+  if [[ "${ROVERD_AUDIO_PLAYBACK_NORMALIZE}" -ne 0 ]]; then
+    ffmpeg_args+=(-af "${ROVERD_AUDIO_PLAYBACK_NORMALIZE_FILTER}")
   fi
 
   ffmpeg_args+=(

pi/bin/audio-only-publisher.sh

@@ -2,26 +2,74 @@
 set -euo pipefail
 set +H
 
-ENV_FILE="${VIDEO_ENV_FILE:-/var/lib/roverd/video.env}"
+ENV_FILE="${ROVERD_MEDIA_ENV_FILE:-/var/lib/roverd/media.env}"
 
 if [[ ! -f "$ENV_FILE" ]]; then
 	echo "Environment file ${ENV_FILE} missing; cannot publish audio" >&2
 	exit 1
 fi
 
-# shellcheck disable=SC1090
-source "$ENV_FILE"
-AUDIO_ENABLE="${AUDIO_ENABLE:-0}"
-if [[ "${AUDIO_ENABLE}" -ne 1 ]]; then
+# Load KEY=VALUE pairs from media.env without evaluating shell syntax. SRT URLs
+# contain characters such as '&' and '#!', so sourcing this file would treat a
+# data file as code and can split a valid URL into shell control operators.
+load_env_file() {
+	local content=""
+
+	if [[ -r "$ENV_FILE" ]]; then
+		content="$(cat "$ENV_FILE")"
+	elif command -v sudo >/dev/null 2>&1; then
+		content="$(sudo -n cat "$ENV_FILE" 2>/dev/null || true)"
+	fi
+
+	if [[ -z "$content" ]]; then
+		echo "Cannot read ${ENV_FILE} (permission denied). Run as a user that can read it, or allow sudo -n for cat." >&2
+		exit 1
+	fi
+
+	local line key val
+	while IFS= read -r line || [[ -n "$line" ]]; do
+		[[ "$line" =~ ^[[:space:]]*$ ]] && continue
+		[[ "$line" =~ ^[[:space:]]*# ]] && continue
+
+		if [[ "$line" =~ ^[[:space:]]*export[[:space:]]+([A-Za-z_][A-Za-z0-9_]*)=(.*)$ ]]; then
+			key="${BASH_REMATCH[1]}"
+			val="${BASH_REMATCH[2]}"
+		elif [[ "$line" =~ ^[[:space:]]*([A-Za-z_][A-Za-z0-9_]*)=(.*)$ ]]; then
+			key="${BASH_REMATCH[1]}"
+			val="${BASH_REMATCH[2]}"
+		else
+			continue
+		fi
+
+		val="${val#${val%%[![:space:]]*}}"
+		val="${val%${val##*[![:space:]]}}"
+		if [[ "$val" =~ ^\".*\"$ ]]; then
+			val="${val:1:${#val}-2}"
+		elif [[ "$val" =~ ^\'.*\'$ ]]; then
+			val="${val:1:${#val}-2}"
+		fi
+
+		printf -v "$key" '%s' "$val"
+		export "$key"
+	done <<< "$content"
+}
+
+load_env_file
+: "${ROVERD_AUDIO_CAPTURE_ENABLE:?ROVERD_AUDIO_CAPTURE_ENABLE not set in ${ENV_FILE}}"
+if [[ "${ROVERD_AUDIO_CAPTURE_ENABLE}" -ne 1 ]]; then
 	echo "Audio capture disabled; skipping audio-only publisher" >&2
 	exit 0
 fi
-: "${AUDIO_PUBLISH_URL:?AUDIO_PUBLISH_URL not set in ${ENV_FILE}}"
+: "${ROVERD_AUDIO_CAPTURE_PUBLISH_URL:?ROVERD_AUDIO_CAPTURE_PUBLISH_URL not set in ${ENV_FILE}}"
+: "${ROVERD_AUDIO_CAPTURE_DEVICE:?ROVERD_AUDIO_CAPTURE_DEVICE not set in ${ENV_FILE}}"
+: "${ROVERD_AUDIO_CAPTURE_SAMPLE_RATE:?ROVERD_AUDIO_CAPTURE_SAMPLE_RATE not set in ${ENV_FILE}}"
+: "${ROVERD_AUDIO_CAPTURE_CHANNELS:?ROVERD_AUDIO_CAPTURE_CHANNELS not set in ${ENV_FILE}}"
+: "${ROVERD_AUDIO_CAPTURE_BITRATE:?ROVERD_AUDIO_CAPTURE_BITRATE not set in ${ENV_FILE}}"
 
-AUDIO_DEVICE="${AUDIO_DEVICE:-hw:0,0}"
+CAPTURE_DEVICE="${ROVERD_AUDIO_CAPTURE_DEVICE}"
 
-# The old 65,536-byte ALSA buffer represented about 171 ms before ffmpeg could
-# even publish the microphone audio:
+# At the default 48 kHz stereo S32_LE capture format, the old 65,536-byte ALSA
+# buffer represented about 171 ms before ffmpeg could publish microphone audio:
 #   65,536 / (48,000 samples * 2 channels * 4 bytes) = 0.1707 seconds
 # Keep the defaults much smaller because this publisher feeds an interactive
 # rover stream, where late-but-smooth audio is less useful than fresher audio.
@@ -51,8 +99,8 @@ run_pipeline() {
 		# format and so the published audio stays full-band stereo before the
 		# Opus encoder sees it.
 		-f s32le
-		-ar 48000
-		-ac 2
+		-ar "${ROVERD_AUDIO_CAPTURE_SAMPLE_RATE}"
+		-ac "${ROVERD_AUDIO_CAPTURE_CHANNELS}"
 		-i pipe:0
 
 		# Keep the known-required microphone boost, but remove the old
@@ -64,9 +112,9 @@ run_pipeline() {
 		# the highest practical quality browsers and MediaMTX can carry without
 		# trying to push raw PCM through the live path.
 		-c:a libopus
-		-b:a 510000
-		-ar:a 48000
-		-ac:a 2
+		-b:a "${ROVERD_AUDIO_CAPTURE_BITRATE}"
+		-ar:a "${ROVERD_AUDIO_CAPTURE_SAMPLE_RATE}"
+		-ac:a "${ROVERD_AUDIO_CAPTURE_CHANNELS}"
 
 		# Use the audio profile for quality. Latency is still controlled by the
 		# 20 ms Opus frame size and the low-buffering capture/publish options
@@ -82,13 +130,13 @@ run_pipeline() {
 		-muxdelay 0
 		-muxpreload 0
 		-f mpegts
-		"${AUDIO_PUBLISH_URL}"
+		"${ROVERD_AUDIO_CAPTURE_PUBLISH_URL}"
 	)
 
-	# The Google Voice HAT microphone path is intentionally fixed instead of
-	# configurable. The previous env-driven sample rate/channel knobs made it
-	# easy for the rover config and the actual ffmpeg pipeline to drift apart,
-	# while the hardware path we install is always 48 kHz stereo capture.
+	# The capture format comes from roverd's normalized media config. Keeping
+	# arecord and ffmpeg on the same env-backed values prevents the publisher
+	# script from quietly disagreeing with roverd.yaml about sample rate or
+	# channel count.
 	#
 	# The buffer and period are byte counts because arecord interprets -B/-F in
 	# microseconds only when the value has an explicit time suffix. Keeping them
@@ -96,7 +144,7 @@ run_pipeline() {
 	# defaults are roughly 43 ms total buffer and 2.7 ms wakeup periods at
 	# 48 kHz stereo S32_LE, which removes about 128 ms of avoidable capture
 	# latency compared with the old 65,536-byte buffer.
-	arecord -D "${AUDIO_DEVICE}" -f S32_LE -c 2 -r 48000 -B "${AUDIO_ALSA_BUFFER_BYTES}" -F "${AUDIO_ALSA_PERIOD_BYTES}" -q -t raw \
+	arecord -D "${CAPTURE_DEVICE}" -f S32_LE -c "${ROVERD_AUDIO_CAPTURE_CHANNELS}" -r "${ROVERD_AUDIO_CAPTURE_SAMPLE_RATE}" -B "${AUDIO_ALSA_BUFFER_BYTES}" -F "${AUDIO_ALSA_PERIOD_BYTES}" -q -t raw \
 		| "${FFMPEG_BIN_PATH}" "${ffmpeg_args[@]}"
 }
 

pi/bin/video-publisher.sh

@@ -4,7 +4,7 @@ set -euo pipefail
 # Keep history expansion off so values containing "!" are safe.
 set +H
 
-ENV_FILE="${VIDEO_ENV_FILE:-/var/lib/roverd/video.env}"
+ENV_FILE="${ROVERD_MEDIA_ENV_FILE:-/var/lib/roverd/media.env}"
 
 if [[ ! -f "$ENV_FILE" ]]; then
   echo "Environment file ${ENV_FILE} missing; cannot publish" >&2
@@ -64,25 +64,29 @@ load_env_file() {
 }
 
 load_env_file
-: "${PUBLISH_URL:?PUBLISH_URL not set in ${ENV_FILE}}"
+: "${ROVERD_VIDEO_ENABLE:?ROVERD_VIDEO_ENABLE not set in ${ENV_FILE}}"
+: "${ROVERD_VIDEO_PUBLISH_URL:?ROVERD_VIDEO_PUBLISH_URL not set in ${ENV_FILE}}"
+: "${ROVERD_VIDEO_WIDTH:?ROVERD_VIDEO_WIDTH not set in ${ENV_FILE}}"
+: "${ROVERD_VIDEO_HEIGHT:?ROVERD_VIDEO_HEIGHT not set in ${ENV_FILE}}"
+: "${ROVERD_VIDEO_FPS:?ROVERD_VIDEO_FPS not set in ${ENV_FILE}}"
+: "${ROVERD_VIDEO_BITRATE:?ROVERD_VIDEO_BITRATE not set in ${ENV_FILE}}"
+: "${ROVERD_VIDEO_INVERT:?ROVERD_VIDEO_INVERT not set in ${ENV_FILE}}"
 
-# Defaults tuned for OV5647: use 4:3 output and force the common 2x2 binned full-FOV mode.
-VIDEO_WIDTH="640"
-VIDEO_HEIGHT="480"
-VIDEO_FPS="30"
-VIDEO_BITRATE="${VIDEO_BITRATE:-3000000}"
-VIDEO_INVERT="${VIDEO_INVERT:-1}"
-VIDEO_SENSOR_MODE="${VIDEO_SENSOR_MODE:-1296:972}"
+if [[ "${ROVERD_VIDEO_ENABLE}" -ne 1 ]]; then
+  echo "Video publisher disabled by roverd media config; skipping" >&2
+  exit 0
+fi
 
-# Flip the camera 180deg by default; allow upright camera mounts via VIDEO_INVERT=0.
+# The inversion decision is made in roverd.yaml and written into media.env, so
+# this script only converts the configured logical value into libcamera flags.
 FLIP_ARGS=()
-if [[ "${VIDEO_INVERT}" -ne 0 ]]; then
+if [[ "${ROVERD_VIDEO_INVERT}" -ne 0 ]]; then
   FLIP_ARGS=(--rotation 180)
 fi
 
 MODE_ARGS=()
-if [[ -n "${VIDEO_SENSOR_MODE}" ]]; then
-  MODE_ARGS=(--mode "${VIDEO_SENSOR_MODE}")
+if [[ -n "${ROVERD_VIDEO_SENSOR_MODE:-}" ]]; then
+  MODE_ARGS=(--mode "${ROVERD_VIDEO_SENSOR_MODE}")
 fi
 
 if [[ -n "${LIBCAMERA_BIN:-}" ]]; then
@@ -110,11 +114,11 @@ run_pipeline() {
     --inline \
     --timeout 0 \
     "${MODE_ARGS[@]}" \
-    --width "${VIDEO_WIDTH}" \
-    --height "${VIDEO_HEIGHT}" \
+    --width "${ROVERD_VIDEO_WIDTH}" \
+    --height "${ROVERD_VIDEO_HEIGHT}" \
     "${FLIP_ARGS[@]}" \
-    --framerate "${VIDEO_FPS}" \
-    --bitrate "${VIDEO_BITRATE}" \
+    --framerate "${ROVERD_VIDEO_FPS}" \
+    --bitrate "${ROVERD_VIDEO_BITRATE}" \
     --codec h264 \
     --profile baseline \
     --denoise auto \
@@ -138,7 +142,7 @@ run_pipeline() {
       -muxdelay 0 \
       -muxpreload 0 \
       -f mpegts \
-      "${PUBLISH_URL}"
+      "${ROVERD_VIDEO_PUBLISH_URL}"
 }
 
 while true; do

pi/install_roverd.sh

@@ -312,18 +312,32 @@ install -D -o root -g root -m 0755 pi/bin/audio-forward-listener.sh /usr/local/b
 install -m 0644 pi/systemd/audio-forward-listener.service /etc/systemd/system/audio-forward-listener.service
 log "Installed audio-forward listener helper + systemd unit"
 install -d -o roverd -g roverd /var/lib/roverd
-cat > /var/lib/roverd/video.env <<'ENV'
+cat > /var/lib/roverd/media.env <<'ENV'
 # Managed by roverd; placeholder values will be overwritten at runtime.
-PUBLISH_URL=srt://192.168.0.86:9000?streamid=#!::r=CHANGE_ME,m=publish&latency=10&mode=caller&transtype=live&pkt_size=1316
-AUDIO_PUBLISH_URL=srt://192.168.0.86:9000?streamid=#!::r=CHANGE_ME-audio,m=publish&latency=10&mode=caller&transtype=live&pkt_size=1316
-AUDIO_FORWARD_URL=srt://192.168.0.86:9000?streamid=#!::r=CHANGE_ME-fwd,m=request&latency=10&mode=caller&transtype=live&pkt_size=1316
-VIDEO_BITRATE=2000000
-AUDIO_ENABLE=0
-AUDIO_DEVICE=hw:0,0
-AUDIO_PLAYBACK_DEVICE=forward
+ROVERD_VIDEO_ENABLE=1
+ROVERD_VIDEO_PUBLISHER=pi-libcamera
+ROVERD_VIDEO_PUBLISH_URL=srt://192.168.0.86:9000?streamid=#!::r=CHANGE_ME,m=publish&latency=10&mode=caller&transtype=live&pkt_size=1316
+ROVERD_VIDEO_DEVICE=
+ROVERD_VIDEO_WIDTH=640
+ROVERD_VIDEO_HEIGHT=480
+ROVERD_VIDEO_FPS=30
+ROVERD_VIDEO_BITRATE=2000000
+ROVERD_VIDEO_INVERT=1
+ROVERD_VIDEO_SENSOR_MODE=1296:972
+ROVERD_AUDIO_CAPTURE_ENABLE=0
+ROVERD_AUDIO_CAPTURE_PUBLISH_URL=srt://192.168.0.86:9000?streamid=#!::r=CHANGE_ME-audio,m=publish&latency=10&mode=caller&transtype=live&pkt_size=1316
+ROVERD_AUDIO_CAPTURE_DEVICE=hw:0,0
+ROVERD_AUDIO_CAPTURE_SAMPLE_RATE=48000
+ROVERD_AUDIO_CAPTURE_CHANNELS=2
+ROVERD_AUDIO_CAPTURE_BITRATE=510000
+ROVERD_AUDIO_PLAYBACK_ENABLE=1
+ROVERD_AUDIO_PLAYBACK_FORWARD_URL=srt://192.168.0.86:9000?streamid=#!::r=CHANGE_ME-fwd,m=request&latency=10&mode=caller&transtype=live&pkt_size=1316
+ROVERD_AUDIO_PLAYBACK_DEVICE=forward
+ROVERD_AUDIO_PLAYBACK_NORMALIZE=1
+ROVERD_AUDIO_PLAYBACK_NORMALIZE_FILTER=dynaudnorm=f=75:g=15:m=10:p=0.9,alimiter=limit=0.85:level=disabled
 ENV
-chown roverd:roverd /var/lib/roverd/video.env
-chmod 0640 /var/lib/roverd/video.env
+chown roverd:roverd /var/lib/roverd/media.env
+chmod 0640 /var/lib/roverd/media.env
 # Create persistent audio FIFO for capture -> publisher
 FIFO_PATH="/var/lib/roverd/audio.pcm"
 if [[ -p "$FIFO_PATH" ]]; then

pi/roverd/cmd/roverd/main.go

@@ -21,7 +21,7 @@ func main() {
 	if err != nil {
 		log.Fatalf("load config: %v", err)
 	}
-	if err := roverd.UpdatePublisherEnv(cfg.Media, cfg.Audio); err != nil {
+	if err := roverd.UpdatePublisherEnv(cfg.Media); err != nil {
 		log.Fatalf("prepare media env: %v", err)
 	}
 
@@ -55,7 +55,7 @@ func main() {
 
 	adapter := roverd.NewSerialAdapter(serialPort, logger)
 
-	mediaSupervisor := roverd.NewMediaSupervisor(cfg.Media, cfg.Audio, logger)
+	mediaSupervisor := roverd.NewMediaSupervisor(cfg.Media, logger)
 	if mediaSupervisor != nil {
 		mediaSupervisor.Start(ctx)
 	}

pi/roverd/config.go

@@ -56,13 +56,13 @@ type BatteryConfig struct {
 }
 
 type AudioConfig struct {
-	CaptureEnabled bool   `yaml:"captureEnabled" json:"captureEnabled"`
-	CaptureDevice  string `yaml:"captureDevice" json:"captureDevice,omitempty"`
-	PlaybackDevice string `yaml:"playbackDevice" json:"playbackDevice,omitempty"`
-	TTSEnabled     bool   `yaml:"ttsEnabled" json:"ttsEnabled"`
-	DefaultEngine  string `yaml:"defaultEngine" json:"defaultEngine,omitempty"`
-	DefaultVoice   string `yaml:"defaultVoice" json:"defaultVoice,omitempty"`
-	DefaultPitch   int    `yaml:"defaultPitch" json:"defaultPitch,omitempty"`
+	// AudioConfig is only for sounds generated by roverd itself. Live capture
+	// and playback are media-publisher concerns, so those settings live under
+	// MediaConfig where they can be written directly into media.env.
+	TTSEnabled    bool   `yaml:"ttsEnabled" json:"ttsEnabled"`
+	DefaultEngine string `yaml:"defaultEngine" json:"defaultEngine,omitempty"`
+	DefaultVoice  string `yaml:"defaultVoice" json:"defaultVoice,omitempty"`
+	DefaultPitch  int    `yaml:"defaultPitch" json:"defaultPitch,omitempty"`
 }
 
 type HornConfig struct {
@@ -77,21 +77,59 @@ type HornConfig struct {
 }
 
 type MediaConfig struct {
-	PublishURL      string   `yaml:"publishUrl" json:"publishUrl,omitempty"`
-	AudioPublishURL string   `yaml:"audioPublishUrl" json:"audioPublishUrl,omitempty"`
-	AudioForwardURL string   `yaml:"audioForwardUrl" json:"audioForwardUrl,omitempty"`
-	PublishPort     int      `yaml:"publishPort" json:"-"`
-	CameraInverted  bool     `yaml:"cameraInverted" json:"-"`
-	Manage          bool     `yaml:"manage"`
-	ManageAudio     bool     `yaml:"manageAudio"`
-	Service         string   `yaml:"service"`
-	AudioService    string   `yaml:"audioService"`
-	HealthURL       string   `yaml:"healthUrl"`
-	HealthInterval  Duration `yaml:"healthInterval"`
-	VideoWidth      int      `yaml:"videoWidth" json:"-"`
-	VideoHeight     int      `yaml:"videoHeight" json:"-"`
-	VideoFPS        int      `yaml:"videoFps" json:"-"`
-	VideoBitrate    int      `yaml:"videoBitrate" json:"-"`
+	// PublishPort is shared by the derived video, microphone, and forwarded-audio
+	// SRT URLs. Keeping it at this level prevents each nested block from needing
+	// to repeat the same server port when the common MediaMTX listener is used.
+	PublishPort    int                 `yaml:"publishPort" json:"-"`
+	Manage         bool                `yaml:"manage" json:"manage"`
+	HealthURL      string              `yaml:"healthUrl" json:"healthUrl,omitempty"`
+	HealthInterval Duration            `yaml:"healthInterval" json:"-"`
+	Video          VideoMediaConfig    `yaml:"video" json:"video"`
+	AudioCapture   AudioCaptureConfig  `yaml:"audioCapture" json:"audioCapture"`
+	AudioPlayback  AudioPlaybackConfig `yaml:"audioPlayback" json:"audioPlayback"`
+}
+
+type VideoMediaConfig struct {
+	// Publisher selects the installed publisher script/pipeline family. The
+	// first pass uses pi-libcamera for current rovers; laptop-v4l2 can be added
+	// without changing the server-facing media shape again.
+	Enabled    bool   `yaml:"enabled" json:"enabled"`
+	Service    string `yaml:"service" json:"service,omitempty"`
+	Publisher  string `yaml:"publisher" json:"publisher,omitempty"`
+	PublishURL string `yaml:"publishUrl" json:"publishUrl,omitempty"`
+	Device     string `yaml:"device" json:"device,omitempty"`
+	Width      int    `yaml:"width" json:"-"`
+	Height     int    `yaml:"height" json:"-"`
+	FPS        int    `yaml:"fps" json:"-"`
+	Bitrate    int    `yaml:"bitrate" json:"-"`
+	Inverted   bool   `yaml:"inverted" json:"-"`
+	SensorMode string `yaml:"sensorMode" json:"-"`
+}
+
+type AudioCaptureConfig struct {
+	// AudioCapture describes the rover microphone stream that browsers can
+	// subscribe to as "<rover>-audio". A disabled capture block still has
+	// normalized defaults so enabling it only requires flipping enabled: true.
+	Enabled    bool   `yaml:"enabled" json:"enabled"`
+	Service    string `yaml:"service" json:"service,omitempty"`
+	PublishURL string `yaml:"publishUrl" json:"publishUrl,omitempty"`
+	Device     string `yaml:"device" json:"device,omitempty"`
+	SampleRate int    `yaml:"sampleRate" json:"-"`
+	Channels   int    `yaml:"channels" json:"-"`
+	Bitrate    int    `yaml:"bitrate" json:"-"`
+}
+
+type AudioPlaybackConfig struct {
+	// AudioPlayback describes the reverse stream that VIP users publish into
+	// MediaMTX for playback on the rover speaker. The URL is a request/read URL
+	// for the rover listener, while the server converts it to publish mode when
+	// it needs to inject audio.
+	Enabled         bool   `yaml:"enabled" json:"enabled"`
+	Service         string `yaml:"service" json:"service,omitempty"`
+	ForwardURL      string `yaml:"forwardUrl" json:"forwardUrl,omitempty"`
+	Device          string `yaml:"device" json:"device,omitempty"`
+	Normalize       bool   `yaml:"normalize" json:"-"`
+	NormalizeFilter string `yaml:"normalizeFilter" json:"-"`
 }
 
 type CameraServoConfig struct {
@@ -189,9 +227,33 @@ func LoadConfig(path string) (*Config, error) {
 		},
 		Media: MediaConfig{
 			PublishPort:    9000,
-			CameraInverted: true,
 			HealthInterval: Duration{Duration: 30 * time.Second},
-			VideoBitrate:   2000000,
+			Video: VideoMediaConfig{
+				Enabled:    true,
+				Service:    "video-publisher.service",
+				Publisher:  "pi-libcamera",
+				Width:      640,
+				Height:     480,
+				FPS:        30,
+				Bitrate:    2000000,
+				Inverted:   true,
+				SensorMode: "1296:972",
+			},
+			AudioCapture: AudioCaptureConfig{
+				Enabled:    false,
+				Service:    "audio-only-publisher.service",
+				Device:     "hw:0,0",
+				SampleRate: 48000,
+				Channels:   2,
+				Bitrate:    510000,
+			},
+			AudioPlayback: AudioPlaybackConfig{
+				Enabled:         true,
+				Service:         "audio-forward-listener.service",
+				Device:          "forward",
+				Normalize:       true,
+				NormalizeFilter: "dynaudnorm=f=75:g=15:m=10:p=0.9,alimiter=limit=0.85:level=disabled",
+			},
 		},
 		CameraServo: CameraServoConfig{
 			Pin:          12,
@@ -205,13 +267,10 @@ func LoadConfig(path string) (*Config, error) {
 			NudgeDegrees: 2,
 		},
 		Audio: AudioConfig{
-			CaptureEnabled: false,
-			CaptureDevice:  "rovermic",
-			PlaybackDevice: "forward",
-			TTSEnabled:     false,
-			DefaultEngine:  "flite",
-			DefaultVoice:   "rms",
-			DefaultPitch:   50,
+			TTSEnabled:    false,
+			DefaultEngine: "flite",
+			DefaultVoice:  "rms",
+			DefaultPitch:  50,
 		},
 		Horn: HornConfig{
 			Enabled:     false,
@@ -283,38 +342,11 @@ func LoadConfig(path string) (*Config, error) {
 	if cfg.BRC.GPIOChip == "" {
 		cfg.BRC.GPIOChip = "gpiochip0"
 	}
-	if cfg.Media.Manage && cfg.Media.Service == "" {
-		return nil, errors.New("media.manage requires media.service")
-	}
-	if cfg.Media.Manage && cfg.Media.HealthInterval.Duration <= 0 {
-		cfg.Media.HealthInterval = Duration{Duration: 30 * time.Second}
-	}
-	if cfg.Media.VideoBitrate <= 0 {
-		cfg.Media.VideoBitrate = 3000000
-	}
 	if cfg.Media.PublishPort <= 0 {
 		cfg.Media.PublishPort = 9000
 	}
-	if cfg.Media.PublishURL == "" {
-		derived, err := derivePublishURL(cfg.ServerURL, cfg.Name, cfg.Media.PublishPort)
-		if err != nil {
-			return nil, fmt.Errorf("derive publishUrl: %w", err)
-		}
-		cfg.Media.PublishURL = derived
-	}
-	if cfg.Media.AudioPublishURL == "" {
-		derived, err := derivePublishURL(cfg.ServerURL, cfg.Name+"-audio", cfg.Media.PublishPort)
-		if err != nil {
-			return nil, fmt.Errorf("derive audioPublishUrl: %w", err)
-		}
-		cfg.Media.AudioPublishURL = derived
-	}
-	if cfg.Media.AudioForwardURL == "" {
-		derived, err := deriveReadURL(cfg.ServerURL, cfg.Name+"-fwd", cfg.Media.PublishPort)
-		if err != nil {
-			return nil, fmt.Errorf("derive audioForwardUrl: %w", err)
-		}
-		cfg.Media.AudioForwardURL = derived
+	if err := validateMediaConfig(&cfg.Media, cfg.ServerURL, cfg.Name); err != nil {
+		return nil, fmt.Errorf("media: %w", err)
 	}
 	if err := validateServoConfig(&cfg.CameraServo); err != nil {
 		return nil, fmt.Errorf("cameraServo: %w", err)
@@ -375,12 +407,6 @@ func clampFloat(value, min, max float64) float64 {
 }
 
 func validateAudioConfig(cfg *AudioConfig) {
-	if cfg.CaptureEnabled && cfg.CaptureDevice == "" {
-		cfg.CaptureDevice = "hw:0,0"
-	}
-	if cfg.PlaybackDevice == "" || cfg.PlaybackDevice == "default" {
-		cfg.PlaybackDevice = "forward"
-	}
 	if cfg.DefaultEngine == "" {
 		cfg.DefaultEngine = "flite"
 	}
@@ -392,6 +418,109 @@ func validateAudioConfig(cfg *AudioConfig) {
 	}
 }
 
+func validateMediaConfig(cfg *MediaConfig, serverURL string, roverName string) error {
+	/*
+		The nested media config is the source of truth for the publisher scripts,
+		so defaults that used to live in shell are normalized here before any env
+		file is written. This keeps the Pi behavior stable while making laptop
+		and future publisher variants explicit configuration choices.
+	*/
+	if cfg.PublishPort <= 0 {
+		cfg.PublishPort = 9000
+	}
+	if cfg.HealthInterval.Duration <= 0 {
+		cfg.HealthInterval = Duration{Duration: 30 * time.Second}
+	}
+	if err := validateVideoMediaConfig(&cfg.Video, serverURL, roverName, cfg.PublishPort); err != nil {
+		return fmt.Errorf("video: %w", err)
+	}
+	if err := validateAudioCaptureConfig(&cfg.AudioCapture, serverURL, roverName, cfg.PublishPort); err != nil {
+		return fmt.Errorf("audioCapture: %w", err)
+	}
+	if err := validateAudioPlaybackConfig(&cfg.AudioPlayback, serverURL, roverName, cfg.PublishPort); err != nil {
+		return fmt.Errorf("audioPlayback: %w", err)
+	}
+	return nil
+}
+
+func validateVideoMediaConfig(cfg *VideoMediaConfig, serverURL string, roverName string, publishPort int) error {
+	if cfg.Service == "" {
+		cfg.Service = "video-publisher.service"
+	}
+	if cfg.Publisher == "" {
+		cfg.Publisher = "pi-libcamera"
+	}
+	if cfg.Width <= 0 {
+		cfg.Width = 640
+	}
+	if cfg.Height <= 0 {
+		cfg.Height = 480
+	}
+	if cfg.FPS <= 0 {
+		cfg.FPS = 30
+	}
+	if cfg.Bitrate <= 0 {
+		cfg.Bitrate = 2000000
+	}
+	if cfg.SensorMode == "" && cfg.Publisher == "pi-libcamera" {
+		cfg.SensorMode = "1296:972"
+	}
+	if cfg.PublishURL == "" {
+		derived, err := derivePublishURL(serverURL, roverName, publishPort)
+		if err != nil {
+			return fmt.Errorf("derive publishUrl: %w", err)
+		}
+		cfg.PublishURL = derived
+	}
+	return nil
+}
+
+func validateAudioCaptureConfig(cfg *AudioCaptureConfig, serverURL string, roverName string, publishPort int) error {
+	if cfg.Service == "" {
+		cfg.Service = "audio-only-publisher.service"
+	}
+	if cfg.Device == "" {
+		cfg.Device = "hw:0,0"
+	}
+	if cfg.SampleRate <= 0 {
+		cfg.SampleRate = 48000
+	}
+	if cfg.Channels <= 0 {
+		cfg.Channels = 2
+	}
+	if cfg.Bitrate <= 0 {
+		cfg.Bitrate = 510000
+	}
+	if cfg.PublishURL == "" {
+		derived, err := derivePublishURL(serverURL, roverName+"-audio", publishPort)
+		if err != nil {
+			return fmt.Errorf("derive publishUrl: %w", err)
+		}
+		cfg.PublishURL = derived
+	}
+	return nil
+}
+
+func validateAudioPlaybackConfig(cfg *AudioPlaybackConfig, serverURL string, roverName string, publishPort int) error {
+	if cfg.Service == "" {
+		cfg.Service = "audio-forward-listener.service"
+	}
+	if cfg.Device == "" {
+		cfg.Device = "forward"
+	}
+	if cfg.NormalizeFilter == "" {
+		cfg.NormalizeFilter = "dynaudnorm=f=75:g=15:m=10:p=0.9,alimiter=limit=0.85:level=disabled"
+	}
+	if cfg.ForwardURL == "" {
+		derived, err := deriveReadURL(serverURL, roverName+"-fwd", publishPort)
+		if err != nil {
+			return fmt.Errorf("derive forwardUrl: %w", err)
+		}
+		cfg.ForwardURL = derived
+	}
+	return nil
+}
+
 func validateHornConfig(cfg *HornConfig) {
 	if cfg.Volume <= 0 {
 		cfg.Volume = 0.25

pi/roverd/media_env.go

@@ -7,51 +7,52 @@ import (
 	"path/filepath"
 )
 
-const publisherEnvPath = "/var/lib/roverd/video.env"
+const publisherEnvPath = "/var/lib/roverd/media.env"
 
-func UpdatePublisherEnv(media MediaConfig, audio AudioConfig) error {
-	if media.PublishURL == "" {
-		return fmt.Errorf("media publishUrl missing")
+func UpdatePublisherEnv(media MediaConfig) error {
+	if media.Video.Enabled && media.Video.PublishURL == "" {
+		return fmt.Errorf("media video publishUrl missing")
 	}
-	if media.AudioPublishURL == "" && audio.CaptureEnabled {
-		return fmt.Errorf("audio publishUrl missing")
+	if media.AudioCapture.Enabled && media.AudioCapture.PublishURL == "" {
+		return fmt.Errorf("media audioCapture publishUrl missing")
 	}
-	if media.VideoWidth < 0 || media.VideoHeight < 0 || media.VideoFPS < 0 || media.VideoBitrate <= 0 {
-		return fmt.Errorf("invalid media dimensions/bitrate")
+	if media.AudioPlayback.Enabled && media.AudioPlayback.ForwardURL == "" {
+		return fmt.Errorf("media audioPlayback forwardUrl missing")
+	}
+	if media.Video.Width <= 0 || media.Video.Height <= 0 || media.Video.FPS <= 0 || media.Video.Bitrate <= 0 {
+		return fmt.Errorf("invalid media video dimensions/bitrate")
 	}
 	if err := os.MkdirAll(filepath.Dir(publisherEnvPath), 0o755); err != nil {
 		return err
 	}
 	var buf bytes.Buffer
-	fmt.Fprintf(&buf, "PUBLISH_URL=%s\n", media.PublishURL)
-	if audio.CaptureEnabled && media.AudioPublishURL != "" {
-		fmt.Fprintf(&buf, "AUDIO_PUBLISH_URL=%s\n", media.AudioPublishURL)
-	}
-	if media.AudioForwardURL != "" {
-		fmt.Fprintf(&buf, "AUDIO_FORWARD_URL=%s\n", media.AudioForwardURL)
-	}
-	if media.VideoWidth > 0 {
-		fmt.Fprintf(&buf, "VIDEO_WIDTH=%d\n", media.VideoWidth)
-	}
-	if media.VideoHeight > 0 {
-		fmt.Fprintf(&buf, "VIDEO_HEIGHT=%d\n", media.VideoHeight)
-	}
-	if media.VideoFPS > 0 {
-		fmt.Fprintf(&buf, "VIDEO_FPS=%d\n", media.VideoFPS)
-	}
-	fmt.Fprintf(&buf, "VIDEO_BITRATE=%d\n", media.VideoBitrate)
-	fmt.Fprintf(&buf, "VIDEO_INVERT=%d\n", boolToInt(media.CameraInverted))
-	audioDevice := audio.CaptureDevice
-	if audioDevice == "" || audioDevice == "rovermic" {
-		audioDevice = "hw:0,0"
-	}
-	fmt.Fprintf(&buf, "AUDIO_ENABLE=%d\n", boolToInt(audio.CaptureEnabled))
-	fmt.Fprintf(&buf, "AUDIO_DEVICE=%s\n", audioDevice)
-	playbackDevice := audio.PlaybackDevice
-	if playbackDevice == "" {
-		playbackDevice = "forward"
-	}
-	fmt.Fprintf(&buf, "AUDIO_PLAYBACK_DEVICE=%s\n", playbackDevice)
+	/*
+		The env file is intentionally verbose: every publisher receives concrete
+		values instead of silently falling back to script-local defaults. That
+		makes roverd.yaml the owner of media behavior while keeping the shell
+		scripts as small pipeline launchers.
+	*/
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_ENABLE=%d\n", boolToInt(media.Video.Enabled))
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_PUBLISHER=%s\n", media.Video.Publisher)
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_PUBLISH_URL=%s\n", media.Video.PublishURL)
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_DEVICE=%s\n", media.Video.Device)
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_WIDTH=%d\n", media.Video.Width)
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_HEIGHT=%d\n", media.Video.Height)
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_FPS=%d\n", media.Video.FPS)
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_BITRATE=%d\n", media.Video.Bitrate)
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_INVERT=%d\n", boolToInt(media.Video.Inverted))
+	fmt.Fprintf(&buf, "ROVERD_VIDEO_SENSOR_MODE=%s\n", media.Video.SensorMode)
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_CAPTURE_ENABLE=%d\n", boolToInt(media.AudioCapture.Enabled))
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_CAPTURE_PUBLISH_URL=%s\n", media.AudioCapture.PublishURL)
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_CAPTURE_DEVICE=%s\n", media.AudioCapture.Device)
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_CAPTURE_SAMPLE_RATE=%d\n", media.AudioCapture.SampleRate)
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_CAPTURE_CHANNELS=%d\n", media.AudioCapture.Channels)
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_CAPTURE_BITRATE=%d\n", media.AudioCapture.Bitrate)
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_PLAYBACK_ENABLE=%d\n", boolToInt(media.AudioPlayback.Enabled))
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_PLAYBACK_FORWARD_URL=%s\n", media.AudioPlayback.ForwardURL)
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_PLAYBACK_DEVICE=%s\n", media.AudioPlayback.Device)
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_PLAYBACK_NORMALIZE=%d\n", boolToInt(media.AudioPlayback.Normalize))
+	fmt.Fprintf(&buf, "ROVERD_AUDIO_PLAYBACK_NORMALIZE_FILTER=%s\n", media.AudioPlayback.NormalizeFilter)
 	if err := os.WriteFile(publisherEnvPath, buf.Bytes(), 0o640); err != nil {
 		return err
 	}

pi/roverd/media_supervisor.go

@@ -13,17 +13,16 @@ import (
 
 type MediaSupervisor struct {
 	cfg           MediaConfig
-	audio         AudioConfig
 	logger        *log.Logger
 	client        *http.Client
 	checkInterval time.Duration
 }
 
-func NewMediaSupervisor(cfg MediaConfig, audio AudioConfig, logger *log.Logger) *MediaSupervisor {
-	if err := UpdatePublisherEnv(cfg, audio); err != nil {
+func NewMediaSupervisor(cfg MediaConfig, logger *log.Logger) *MediaSupervisor {
+	if err := UpdatePublisherEnv(cfg); err != nil {
 		logger.Printf("media supervisor: update env failed: %v", err)
 	}
-	if !cfg.Manage || cfg.Service == "" {
+	if !cfg.Manage {
 		return nil
 	}
 	interval := cfg.HealthInterval.Duration
@@ -36,7 +35,6 @@ func NewMediaSupervisor(cfg MediaConfig, audio AudioConfig, logger *log.Logger)
 	}
 	return &MediaSupervisor{
 		cfg:           cfg,
-		audio:         audio,
 		logger:        logger,
 		client:        client,
 		checkInterval: interval,
@@ -47,7 +45,7 @@ func (m *MediaSupervisor) Start(ctx context.Context) {
 	if m == nil {
 		return
 	}
-	if err := UpdatePublisherEnv(m.cfg, m.audio); err != nil {
+	if err := UpdatePublisherEnv(m.cfg); err != nil {
 		m.logger.Printf("media supervisor: update env failed: %v", err)
 	}
 	if m.cfg.HealthURL == "" || m.client == nil {
@@ -78,7 +76,7 @@ func (m *MediaSupervisor) HandleAction(ctx context.Context, action string) error
 	if m == nil {
 		return errors.New("media supervisor disabled")
 	}
-	if err := UpdatePublisherEnv(m.cfg, m.audio); err != nil {
+	if err := UpdatePublisherEnv(m.cfg); err != nil {
 		return err
 	}
 	switch action {
@@ -96,9 +94,9 @@ func (m *MediaSupervisor) checkAndRepair() error {
 	if m.checkHealth(ctx) {
 		return nil
 	}
-	m.logger.Printf("media supervisor: health check failed, restarting %s", m.cfg.Service)
+	m.logger.Printf("media supervisor: health check failed, restarting configured media services")
 	if err := m.runSystemctl(ctx, "restart"); err != nil {
-		return fmt.Errorf("restart mediamtx: %w", err)
+		return fmt.Errorf("restart media services: %w", err)
 	}
 	return nil
 }
@@ -127,16 +125,39 @@ func (m *MediaSupervisor) checkHealth(ctx context.Context) bool {
 }
 
 func (m *MediaSupervisor) runSystemctl(ctx context.Context, action string) error {
-	if m.cfg.Service == "" {
-		return errors.New("no media service configured")
+	services := m.managedServices()
+	if len(services) == 0 {
+		return errors.New("no media services configured")
 	}
 	runCtx, cancel := context.WithTimeout(ctx, 15*time.Second)
 	defer cancel()
 
-	cmd := exec.CommandContext(runCtx, "systemctl", action, m.cfg.Service)
-	output, err := cmd.CombinedOutput()
-	if err != nil {
-		return fmt.Errorf("systemctl %s %s: %w (%s)", action, m.cfg.Service, err, string(output))
+	for _, service := range services {
+		cmd := exec.CommandContext(runCtx, "systemctl", action, service)
+		output, err := cmd.CombinedOutput()
+		if err != nil {
+			return fmt.Errorf("systemctl %s %s: %w (%s)", action, service, err, string(output))
+		}
 	}
 	return nil
 }
+
+func (m *MediaSupervisor) managedServices() []string {
+	/*
+		Only enabled streams are service-managed. This matters for audio capture:
+		the service may be installed everywhere, but a rover with capture disabled
+		should not have remote media commands starting a publisher that the config
+		says should stay off.
+	*/
+	services := []string{}
+	if m.cfg.Video.Enabled && m.cfg.Video.Service != "" {
+		services = append(services, m.cfg.Video.Service)
+	}
+	if m.cfg.AudioCapture.Enabled && m.cfg.AudioCapture.Service != "" {
+		services = append(services, m.cfg.AudioCapture.Service)
+	}
+	if m.cfg.AudioPlayback.Enabled && m.cfg.AudioPlayback.Service != "" {
+		services = append(services, m.cfg.AudioPlayback.Service)
+	}
+	return services
+}

pi/roverd/roverd.sample.yaml

@@ -17,16 +17,36 @@ battery:
   urgent: 1650
 maxWheelSpeed: 350
 media:
-  publishUrl: srt://192.168.0.86:9000?streamid=#!::r=roomba-alpha,m=publish&latency=10&mode=caller&transtype=live&pkt_size=1316
-  audioForwardUrl: srt://192.168.0.86:9000?streamid=#!::r=roomba-alpha-fwd,m=request&latency=10&mode=caller&transtype=live&pkt_size=1316
   publishPort: 9000
-  # Default assumes camera is mounted upside down; set false for upright mounts.
-  cameraInverted: true
-  videoBitrate: 2000000
   manage: true
-  service: video-publisher.service
   healthUrl: ""
   healthInterval: 30s
+  video:
+    enabled: true
+    service: video-publisher.service
+    publisher: pi-libcamera
+    publishUrl: srt://192.168.0.86:9000?streamid=#!::r=roomba-alpha,m=publish&latency=10&mode=caller&transtype=live&pkt_size=1316
+    width: 640
+    height: 480
+    fps: 30
+    bitrate: 2000000
+    # Default assumes camera is mounted upside down; set false for upright mounts.
+    inverted: true
+    sensorMode: 1296:972
+  audioCapture:
+    enabled: false
+    service: audio-only-publisher.service
+    publishUrl: srt://192.168.0.86:9000?streamid=#!::r=roomba-alpha-audio,m=publish&latency=10&mode=caller&transtype=live&pkt_size=1316
+    device: hw:0,0
+    sampleRate: 48000
+    channels: 2
+    bitrate: 510000
+  audioPlayback:
+    enabled: true
+    service: audio-forward-listener.service
+    forwardUrl: srt://192.168.0.86:9000?streamid=#!::r=roomba-alpha-fwd,m=request&latency=10&mode=caller&transtype=live&pkt_size=1316
+    device: forward
+    normalize: true
 cameraServo:
   enabled: false
   pin: 12

pi/roverd/roverd.yaml

@@ -17,9 +17,18 @@ battery:
 maxWheelSpeed: 350
 media:
   manage: false
-  service: mediamtx.service
   healthUrl: http://127.0.0.1:9997/v3/paths/list
   healthInterval: 30s
+  video:
+    enabled: true
+    service: video-publisher.service
+    publisher: pi-libcamera
+  audioCapture:
+    enabled: false
+    service: audio-only-publisher.service
+  audioPlayback:
+    enabled: true
+    service: audio-forward-listener.service
 cameraServo:
   enabled: false
   pin: 19

pi/systemd/audio-forward-listener.service

@@ -7,7 +7,7 @@ Wants=network-online.target
 Type=simple
 User=roverd
 Group=roverd
-EnvironmentFile=/var/lib/roverd/video.env
+EnvironmentFile=/var/lib/roverd/media.env
 ExecStart=/usr/local/bin/audio-forward-listener
 KillMode=control-group
 TimeoutStopSec=5

pi/systemd/audio-only-publisher.service

@@ -7,7 +7,7 @@ Wants=network-online.target
 Type=simple
 User=roverd
 Group=roverd
-EnvironmentFile=/var/lib/roverd/video.env
+EnvironmentFile=/var/lib/roverd/media.env
 ExecStart=/usr/local/bin/audio-only-publisher
 KillMode=control-group
 TimeoutStopSec=5

pi/systemd/video-publisher.service

@@ -8,7 +8,7 @@ Type=simple
 User=roverd
 Group=roverd
 WorkingDirectory=/var/lib/roverd
-EnvironmentFile=/var/lib/roverd/video.env
+EnvironmentFile=/var/lib/roverd/media.env
 ExecStart=/usr/local/bin/video-publisher
 Restart=always
 RestartSec=2

server/public/index.html

@@ -78,7 +78,7 @@
 <script defer src="https://analytics.otter.land/script.js" data-website-id="82dd56a5-db44-4279-bd1e-a4d9fee39af7" data-domains="rover.otter.land"></script>
 <script defer src="https://analytics.otter.land/recorder.js" data-website-id="82dd56a5-db44-4279-bd1e-a4d9fee39af7" data-domains="rover.otter.land" data-sample-rate="0.15" data-mask-level="moderate" data-max-duration="300000"></script>
     <title>Roomba Rover</title>
-    <script type="module" crossorigin src="/assets/index-yGP4eLx4.js"></script>
+    <script type="module" crossorigin src="/assets/index-DRqsCa1W.js"></script>
     <link rel="stylesheet" crossorigin href="/assets/index-JqEX_oga.css">
   </head>
   <body>

server/src/services/audioForwardService/policy.js

@@ -37,7 +37,10 @@ function createAudioForwardPolicy(deps) {
 
   function resolveForwardUrl(roverId) {
     const record = roverManager.rovers.get(roverId);
-    const configured = record?.meta?.media?.audioForwardUrl;
+    // Rovers listen to the playback stream with a request/read URL. The VIP
+    // upload path needs to publish into that same stream, so the configured
+    // nested playback URL is converted to publish mode below.
+    const configured = record?.meta?.media?.audioPlayback?.forwardUrl;
     if (configured) return forcePublishStreamMode(configured);
     return `srt://127.0.0.1:9000?streamid=#!::r=${encodeURIComponent(
       roverId + streamSuffix,

server/src/services/replayEngineV2/sources.js

@@ -25,12 +25,19 @@ function getRoomCameraStream(camera) {
   return null;
 }
 
+function hasRoverAudioCapture(rover) {
+  // The replay worker only records a rover microphone stream when roverd says
+  // capture is both enabled and publishable. That mirrors the media publisher
+  // contract instead of guessing from stream naming conventions alone.
+  return Boolean(rover?.media?.audioCapture?.enabled && rover?.media?.audioCapture?.publishUrl);
+}
+
 function listDesiredSources() {
   const sources = [];
   for (const rover of roverManager.getRoster()) {
     const roverId = String(rover.id);
     sources.push({ id: roverId, sourceType: 'rover', kind: 'video', label: rover.name || roverId, inputUrl: toSrtReadPath(roverId) });
-    if (rover?.media?.audioPublishUrl) {
+    if (hasRoverAudioCapture(rover)) {
       sources.push({ id: `${roverId}-audio`, sourceType: 'rover', roverId, kind: 'audio', label: `${rover.name || roverId} audio`, inputUrl: toSrtReadPath(`${roverId}-audio`) });
     }
   }

webui/src/components/RoverMediaPlayer/index.jsx

@@ -40,6 +40,14 @@ function normalizeVideoFilter(value) {
     : VIDEO_SETTINGS_DEFAULTS.colorFilter;
 }
 
+function hasRoverAudioCapture(rover) {
+  // The browser asks for the separate "<rover>-audio" WHEP source only when
+  // roverd's media config says the microphone publisher is enabled and has a
+  // concrete publish URL. That keeps UI media setup aligned with the daemon
+  // config instead of probing for streams that should not exist.
+  return Boolean(rover?.media?.audioCapture?.enabled && rover?.media?.audioCapture?.publishUrl);
+}
+
 export default function RoverMediaPlayer({
   roverId = null,
   sessionInfo = null,
@@ -58,7 +66,7 @@ export default function RoverMediaPlayer({
       ? state.session.roster.find((item) => String(item.id) === String(effectiveRoverId)) || null
       : null,
   );
-  const hasAudio = Boolean(rosterEntry?.media?.audioPublishUrl);
+  const hasAudio = hasRoverAudioCapture(rosterEntry);
   const autoVideoEnabled = videoMode ? videoMode === 'whep' : true;
   const autoAudioEnabled = hasAudio;
   const autoEntries = useMemo(() => {

webui/src/mini/MiniSummaryApp/MiniSummaryContent.jsx

@@ -15,6 +15,12 @@ import InfoColumn from './components/InfoColumn.jsx';
 import { ROTATE_MS } from './constants.js';
 import { formatDriverLabel } from './utils.js';
 
+function hasRoverAudioCapture(rover) {
+  // Mini uses the same media contract as the full rover player: audio exists
+  // only when the nested microphone publisher block is enabled and publishable.
+  return Boolean(rover?.media?.audioCapture?.enabled && rover?.media?.audioCapture?.publishUrl);
+}
+
 export default function MiniSummaryContent() {
   const { session } = useSession();
   const spectatorReady = useSpectatorMode();
@@ -53,7 +59,7 @@ export default function MiniSummaryContent() {
   const audioEntries = useMemo(
     () =>
       driverRoster.flatMap((rover) => {
-        if (!rover?.id || !rover.media?.audioPublishUrl) return [];
+        if (!rover?.id || !hasRoverAudioCapture(rover)) return [];
         const id = String(rover.id);
         return [{ type: 'rover', id: `${id}-audio`, key: `${id}-audio` }];
       }),