Record replay audio

images/chromium-headful/Dockerfile

@@ -366,8 +366,8 @@ COPY --from=xorg-deps /usr/local/lib/xorg/modules/drivers/dummy_drv.so /usr/lib/
 COPY --from=xorg-deps /usr/local/lib/xorg/modules/input/neko_drv.so /usr/lib/xorg/modules/input/neko_drv.so
 
 COPY images/chromium-headful/image-chromium/ /
-COPY images/chromium-headful/start-pulseaudio.sh /images/chromium-headful/start-pulseaudio.sh
-RUN chmod +x /images/chromium-headful/start-pulseaudio.sh
+COPY shared/start-pulseaudio.sh /usr/local/bin/start-pulseaudio.sh
+RUN chmod +x /usr/local/bin/start-pulseaudio.sh
 COPY images/chromium-headful/supervisord.conf /etc/supervisor/supervisord.conf
 COPY images/chromium-headful/supervisor/services/ /etc/supervisor/conf.d/services/
 COPY shared/envoy/supervisor-envoy.conf /etc/supervisor/conf.d/services/envoy.conf

images/chromium-headful/neko.yaml

@@ -42,6 +42,14 @@ session:
     # needed for legacy API
     enabled: false
 
+capture:
+  audio:
+    # Capture from the monitor of the recorder's playback sink so live view
+    # streams the same browser audio that gets recorded. Neko defaults to
+    # "audio_output.monitor", which does not exist here (our sink is
+    # KernelOutput), so without this override live view has no audio.
+    device: "KernelOutput.monitor"
+
 plugins:
   enabled: false
 

images/chromium-headful/supervisor/services/kernel-images-api.conf

@@ -1,5 +1,5 @@
 [program:kernel-images-api]
-command=/bin/bash -lc 'mkdir -p "${KERNEL_IMAGES_API_OUTPUT_DIR:-/recordings}" && PORT="${KERNEL_IMAGES_API_PORT:-10001}" FRAME_RATE="${KERNEL_IMAGES_API_FRAME_RATE:-10}" DISPLAY_NUM="${KERNEL_IMAGES_API_DISPLAY_NUM:-${DISPLAY_NUM:-1}}" MAX_SIZE_MB="${KERNEL_IMAGES_API_MAX_SIZE_MB:-500}" OUTPUT_DIR="${KERNEL_IMAGES_API_OUTPUT_DIR:-/recordings}" LOG_CDP_MESSAGES="${LOG_CDP_MESSAGES:-false}" S2_BASIN="${S2_BASIN:-}" S2_ACCESS_TOKEN="${S2_ACCESS_TOKEN:-}" S2_STREAM="${S2_STREAM:-}" exec /usr/local/bin/kernel-images-api'
+command=/bin/bash -lc 'mkdir -p "${KERNEL_IMAGES_API_OUTPUT_DIR:-/recordings}" && PORT="${KERNEL_IMAGES_API_PORT:-10001}" FRAME_RATE="${KERNEL_IMAGES_API_FRAME_RATE:-10}" DISPLAY_NUM="${KERNEL_IMAGES_API_DISPLAY_NUM:-${DISPLAY_NUM:-1}}" MAX_SIZE_MB="${KERNEL_IMAGES_API_MAX_SIZE_MB:-500}" OUTPUT_DIR="${KERNEL_IMAGES_API_OUTPUT_DIR:-/recordings}" RECORD_AUDIO="${KERNEL_IMAGES_API_RECORD_AUDIO:-${RECORD_AUDIO:-true}}" AUDIO_SOURCE="${KERNEL_IMAGES_API_AUDIO_SOURCE:-${AUDIO_SOURCE:-KernelOutput.monitor}}" PULSE_SERVER="${PULSE_SERVER:-unix:/tmp/pulse/native}" PULSE_SINK="${PULSE_SINK:-KernelOutput}" LOG_CDP_MESSAGES="${LOG_CDP_MESSAGES:-false}" S2_BASIN="${S2_BASIN:-}" S2_ACCESS_TOKEN="${S2_ACCESS_TOKEN:-}" S2_STREAM="${S2_STREAM:-}" exec /usr/local/bin/kernel-images-api'
 autostart=false
 autorestart=true
 startsecs=0

images/chromium-headful/supervisor/services/pulseaudio.conf

@@ -1,7 +1,7 @@
 [program:pulseaudio]
-command=/bin/bash -lc '/images/chromium-headful/start-pulseaudio.sh'
+command=/bin/bash -lc '/usr/local/bin/start-pulseaudio.sh'
 autostart=false
-autorestart=false
+autorestart=true
 startsecs=0
 exitcodes=0
 stdout_logfile=/var/log/supervisord/pulseaudio

images/chromium-headless/image/Dockerfile

@@ -149,6 +149,7 @@ RUN --mount=type=cache,target=/var/cache/apt,sharing=locked,id=$CACHEIDPREFIX-ap
     gpg-agent \
     dbus \
     dbus-x11 \
+    pulseaudio \
     xvfb \
     x11-utils \
     x11-xserver-utils \
@@ -235,6 +236,8 @@ RUN useradd -m -s /bin/bash kernel
 # supervisor  start scripts
 COPY images/chromium-headless/image/start-xvfb.sh /images/chromium-headless/image/start-xvfb.sh
 RUN chmod +x /images/chromium-headless/image/start-xvfb.sh
+COPY shared/start-pulseaudio.sh /usr/local/bin/start-pulseaudio.sh
+RUN chmod +x /usr/local/bin/start-pulseaudio.sh
 
 # Container entrypoint wrapper (Go binary, replaces wrapper.sh)
 COPY --from=server-builder /out/wrapper /wrapper

images/chromium-headless/image/supervisor/services/kernel-images-api.conf

@@ -1,5 +1,5 @@
 [program:kernel-images-api]
-command=/bin/bash -lc 'mkdir -p "${KERNEL_IMAGES_API_OUTPUT_DIR:-/recordings}" && PORT="${KERNEL_IMAGES_API_PORT:-10001}" FRAME_RATE="${KERNEL_IMAGES_API_FRAME_RATE:-10}" DISPLAY_NUM="${KERNEL_IMAGES_API_DISPLAY_NUM:-${DISPLAY_NUM:-1}}" MAX_SIZE_MB="${KERNEL_IMAGES_API_MAX_SIZE_MB:-500}" OUTPUT_DIR="${KERNEL_IMAGES_API_OUTPUT_DIR:-/recordings}" LOG_CDP_MESSAGES="${LOG_CDP_MESSAGES:-false}" S2_BASIN="${S2_BASIN:-}" S2_ACCESS_TOKEN="${S2_ACCESS_TOKEN:-}" S2_STREAM="${S2_STREAM:-}" exec /usr/local/bin/kernel-images-api'
+command=/bin/bash -lc 'mkdir -p "${KERNEL_IMAGES_API_OUTPUT_DIR:-/recordings}" && PORT="${KERNEL_IMAGES_API_PORT:-10001}" FRAME_RATE="${KERNEL_IMAGES_API_FRAME_RATE:-10}" DISPLAY_NUM="${KERNEL_IMAGES_API_DISPLAY_NUM:-${DISPLAY_NUM:-1}}" MAX_SIZE_MB="${KERNEL_IMAGES_API_MAX_SIZE_MB:-500}" OUTPUT_DIR="${KERNEL_IMAGES_API_OUTPUT_DIR:-/recordings}" RECORD_AUDIO="${KERNEL_IMAGES_API_RECORD_AUDIO:-${RECORD_AUDIO:-true}}" AUDIO_SOURCE="${KERNEL_IMAGES_API_AUDIO_SOURCE:-${AUDIO_SOURCE:-KernelOutput.monitor}}" PULSE_SERVER="${PULSE_SERVER:-unix:/tmp/pulse/native}" PULSE_SINK="${PULSE_SINK:-KernelOutput}" LOG_CDP_MESSAGES="${LOG_CDP_MESSAGES:-false}" S2_BASIN="${S2_BASIN:-}" S2_ACCESS_TOKEN="${S2_ACCESS_TOKEN:-}" S2_STREAM="${S2_STREAM:-}" exec /usr/local/bin/kernel-images-api'
 autostart=false
 autorestart=true
 startsecs=0

images/chromium-headless/image/supervisor/services/pulseaudio.conf

@@ -0,0 +1,8 @@
+[program:pulseaudio]
+command=/bin/bash -lc '/usr/local/bin/start-pulseaudio.sh'
+autostart=false
+autorestart=true
+startsecs=0
+exitcodes=0
+stdout_logfile=/var/log/supervisord/pulseaudio
+redirect_stderr=true

server/cmd/api/main.go

@@ -72,6 +72,8 @@ func main() {
 		FrameRate:   &config.FrameRate,
 		MaxSizeInMB: &config.MaxSizeInMB,
 		OutputDir:   &config.OutputDir,
+		RecordAudio: &config.RecordAudio,
+		AudioSource: &config.AudioSource,
 	}
 	if err := defaultParams.Validate(); err != nil {
 		slogger.Error("invalid default recording parameters", "err", err)

server/cmd/chromium-launcher/main.go

@@ -122,6 +122,8 @@ func main() {
 		"env",
 		"DISPLAY=:1",
 		"DBUS_SESSION_BUS_ADDRESS=unix:path=/run/dbus/system_bus_socket",
+		"PULSE_SERVER=" + os.Getenv("PULSE_SERVER"),
+		"PULSE_SINK=" + os.Getenv("PULSE_SINK"),
 		"XDG_CONFIG_HOME=/home/kernel/.config",
 		"XDG_CACHE_HOME=/home/kernel/.cache",
 		"HOME=/home/kernel",

server/cmd/config/config.go

@@ -18,6 +18,8 @@ type Config struct {
 	DisplayNum  int    `envconfig:"DISPLAY_NUM" default:"1"`
 	MaxSizeInMB int    `envconfig:"MAX_SIZE_MB" default:"500"`
 	OutputDir   string `envconfig:"OUTPUT_DIR" default:"."`
+	RecordAudio bool   `envconfig:"RECORD_AUDIO" default:"false"`
+	AudioSource string `envconfig:"AUDIO_SOURCE" default:"KernelOutput.monitor"`
 
 	// Absolute or relative path to the ffmpeg binary. If empty the code falls back to "ffmpeg" on $PATH.
 	PathToFFmpeg string `envconfig:"FFMPEG_PATH" default:"ffmpeg"`
@@ -55,6 +57,8 @@ func (c *Config) LogValue() slog.Value {
 		slog.Int("display_num", c.DisplayNum),
 		slog.Int("max_size_mb", c.MaxSizeInMB),
 		slog.String("output_dir", c.OutputDir),
+		slog.Bool("record_audio", c.RecordAudio),
+		slog.String("audio_source", c.AudioSource),
 		slog.String("ffmpeg_path", c.PathToFFmpeg),
 		slog.Int("devtools_proxy_port", c.DevToolsProxyPort),
 		slog.Bool("log_cdp_messages", c.LogCDPMessages),
@@ -100,6 +104,9 @@ func validate(config *Config) error {
 	if config.PathToFFmpeg == "" {
 		return fmt.Errorf("FFMPEG_PATH is required")
 	}
+	if config.RecordAudio && config.AudioSource == "" {
+		return fmt.Errorf("AUDIO_SOURCE is required when RECORD_AUDIO is true")
+	}
 	if config.ChromeDriverUpstreamAddr == "" {
 		return fmt.Errorf("CHROMEDRIVER_UPSTREAM_ADDR is required")
 	}

server/cmd/config/config_test.go

@@ -23,6 +23,8 @@ func TestLoad(t *testing.T) {
 				DisplayNum:               1,
 				MaxSizeInMB:              500,
 				OutputDir:                ".",
+				RecordAudio:              false,
+				AudioSource:              "KernelOutput.monitor",
 				PathToFFmpeg:             "ffmpeg",
 				DevToolsProxyPort:        9222,
 				ScaleToZeroCooldown:      time.Second,
@@ -39,6 +41,8 @@ func TestLoad(t *testing.T) {
 				"DISPLAY_NUM":                "2",
 				"MAX_SIZE_MB":                "250",
 				"OUTPUT_DIR":                 "/tmp",
+				"RECORD_AUDIO":               "true",
+				"AUDIO_SOURCE":               "CustomOutput.monitor",
 				"FFMPEG_PATH":                "/usr/local/bin/ffmpeg",
 				"DEVTOOLS_PROXY_PORT":        "9876",
 				"SCALE_TO_ZERO_COOLDOWN":     "5s",
@@ -51,6 +55,8 @@ func TestLoad(t *testing.T) {
 				DisplayNum:               2,
 				MaxSizeInMB:              250,
 				OutputDir:                "/tmp",
+				RecordAudio:              true,
+				AudioSource:              "CustomOutput.monitor",
 				PathToFFmpeg:             "/usr/local/bin/ffmpeg",
 				DevToolsProxyPort:        9876,
 				ScaleToZeroCooldown:      5 * time.Second,
@@ -71,6 +77,8 @@ func TestLoad(t *testing.T) {
 				DisplayNum:               1,
 				MaxSizeInMB:              500,
 				OutputDir:                ".",
+				RecordAudio:              false,
+				AudioSource:              "KernelOutput.monitor",
 				PathToFFmpeg:             "ffmpeg",
 				DevToolsProxyPort:        7777,
 				ScaleToZeroCooldown:      time.Second,
@@ -114,6 +122,14 @@ func TestLoad(t *testing.T) {
 			},
 			wantErr: true,
 		},
+		{
+			name: "missing audio source when recording audio",
+			env: map[string]string{
+				"RECORD_AUDIO": "true",
+				"AUDIO_SOURCE": "",
+			},
+			wantErr: true,
+		},
 		{
 			name: "missing chromedriver upstream addr (set to empty)",
 			env: map[string]string{

server/cmd/wrapper/chromium.go

@@ -47,7 +47,6 @@ func applyHeadlessDefaultFlags() {
 		"--hide-crash-restore-bubble",
 		"--hide-scrollbars",
 		"--metrics-recording-only",
-		"--mute-audio",
 		"--no-default-browser-check",
 		"--no-first-run",
 		"--no-sandbox",

server/cmd/wrapper/main.go

@@ -27,6 +27,7 @@ const (
 	dbusSocket     = "/run/dbus/system_bus_socket"
 	defaultDisplay = ":1"
 	defaultIntPort = "9223"
+	pulseSocket    = "/tmp/pulse/native"
 )
 
 type profile int
@@ -119,6 +120,15 @@ func main() {
 	// starts so it captures the env for child services (notably chromium,
 	// which would otherwise spam autolaunch errors).
 	_ = os.Setenv("DBUS_SESSION_BUS_ADDRESS", "unix:path="+dbusSocket)
+	if os.Getenv("PULSE_SERVER") == "" {
+		_ = os.Setenv("PULSE_SERVER", "unix:"+pulseSocket)
+	}
+	if os.Getenv("PULSE_SINK") == "" {
+		_ = os.Setenv("PULSE_SINK", "KernelOutput")
+	}
+	if os.Getenv("AUDIO_SOURCE") == "" {
+		_ = os.Setenv("AUDIO_SOURCE", "KernelOutput.monitor")
+	}
 
 	// Stale X locks from prior runs.
 	_ = os.Remove("/tmp/.X1-lock")
@@ -172,11 +182,13 @@ func main() {
 	_ = os.WriteFile(filepath.Join(supervisordLogD, "chromium"), nil, 0o644)
 
 	browserStart := time.Now()
-	startAll(xServer, "dbus", "chromedriver", "chromium")
+	startAll(xServer, "dbus", "chromedriver", "pulseaudio")
 	waitForX(defaultDisplay, 20*time.Second)
 	if prof == profileHeadful {
 		startAll("mutter")
 	}
+	waitForSocket(pulseSocket, 10*time.Second)
+	startAll("chromium")
 	waitForSocket(dbusSocket, 10*time.Second)
 	if prof == profileHeadful && webrtc {
 		startAll("neko")
@@ -223,12 +235,6 @@ func main() {
 		identityDone.Sub(identityStart).Truncate(time.Millisecond),
 		formatProbeDurations(probeDurations))
 
-	// Cosmetic + non-critical services come up off the hot path. Headless has
-	// no audio stack.
-	if prof == profileHeadful {
-		go startAll("pulseaudio")
-	}
-
 	// Re-enable scale-to-zero now that the hot path is up — unless the caller
 	// asked to keep it disabled via ENABLE_STZ=false/0.
 	if stzManaged {