console: disable agent audio I/O in text mode (#845)

Author: theomonnom

cmd/lk/console_tui.go

@@ -151,6 +151,7 @@ func (m consoleModel) Init() tea.Cmd {
 	}
 	if m.textMode {
 		cmds = append(cmds, textinput.Blink)
+		m.applyTextMode(true)
 	}
 	return tea.Batch(cmds...)
 }
@@ -216,6 +217,7 @@ func (m consoleModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 			m.showShortcuts = false
 			m.partialTranscript = ""
 			m.textInput.Focus()
+			m.applyTextMode(true)
 			return m, textinput.Blink
 		case "?":
 			m.showShortcuts = !m.showShortcuts
@@ -259,6 +261,7 @@ func (m consoleModel) Update(msg tea.Msg) (tea.Model, tea.Cmd) {
 			m.audioError = ""
 			m.inputDev = "Default Input"
 			m.outputDev = "Default Output"
+			m.applyTextMode(false)
 		}
 		return m, nil
 
@@ -289,6 +292,7 @@ func (m *consoleModel) switchToAudio() tea.Cmd {
 		m.showShortcuts = false
 		m.textInput.Blur()
 		m.audioError = ""
+		m.applyTextMode(false)
 		return nil
 	}
 	// Lazy init audio in a goroutine
@@ -297,6 +301,34 @@ func (m *consoleModel) switchToAudio() tea.Cmd {
 	}
 }
 
+// applyTextMode pauses the local audio pipeline and asks the agent to
+// disable/enable audio I/O so STT/TTS aren't running in text mode.
+func (m *consoleModel) applyTextMode(text bool) {
+	if m.pipeline.HasAudio() {
+		m.pipeline.SetPaused(text)
+	}
+
+	m.reqCounter++
+	reqID := fmt.Sprintf("console-io-%d", m.reqCounter)
+	audioOn := !text
+	transcriptionOn := !text
+	req := &agent.SessionRequest{
+		RequestId: reqID,
+		Request: &agent.SessionRequest_UpdateIo{
+			UpdateIo: &agent.SessionRequest_UpdateIO{
+				Input: &agent.SessionRequest_UpdateIO_Input{
+					AudioEnabled: &audioOn,
+				},
+				Output: &agent.SessionRequest_UpdateIO_Output{
+					AudioEnabled:         &audioOn,
+					TranscriptionEnabled: &transcriptionOn,
+				},
+			},
+		},
+	}
+	go m.pipeline.SendRequest(req)
+}
+
 func (m *consoleModel) beginShutdown() tea.Cmd {
 	m.shuttingDown = true
 	m.textMode = false

go.mod

@@ -18,7 +18,7 @@ require (
 	github.com/google/go-containerregistry v0.20.6
 	github.com/google/go-querystring v1.2.0
 	github.com/joho/godotenv v1.5.1
-	github.com/livekit/protocol v1.45.9-0.20260508203311-a249893d6a5d
+	github.com/livekit/protocol v1.45.9-0.20260519220930-d0b708548748
 	github.com/livekit/server-sdk-go/v2 v2.16.2
 	github.com/mattn/go-isatty v0.0.21
 	github.com/moby/patternmatcher v0.6.1

go.sum

@@ -366,8 +366,8 @@ github.com/livekit/mageutil v0.0.0-20250511045019-0f1ff63f7731 h1:9x+U2HGLrSw5AT
 github.com/livekit/mageutil v0.0.0-20250511045019-0f1ff63f7731/go.mod h1:Rs3MhFwutWhGwmY1VQsygw28z5bWcnEYmS1OG9OxjOQ=
 github.com/livekit/mediatransportutil v0.0.0-20260309115634-0e2e24b36ee8 h1:coWig9fKxdb/nwOaIoGUUAogso12GblAJh/9SA9hcxk=
 github.com/livekit/mediatransportutil v0.0.0-20260309115634-0e2e24b36ee8/go.mod h1:RCd46PT+6sEztld6XpkCrG1xskb0u3SqxIjy4G897Ss=
-github.com/livekit/protocol v1.45.9-0.20260508203311-a249893d6a5d h1:mE0/AjgGnvsF3q0ipiNGe1HZ3CUKUG7Y+zqey/2JrBE=
-github.com/livekit/protocol v1.45.9-0.20260508203311-a249893d6a5d/go.mod h1:KEPIJ/ZdMFQ9tmmfv/uT9TjQEuEcZupCZBabuRGEC1k=
+github.com/livekit/protocol v1.45.9-0.20260519220930-d0b708548748 h1:QXPwHAEWREHebK+NW+zdgz4x/Dw+I4U5WxVeEbZolwo=
+github.com/livekit/protocol v1.45.9-0.20260519220930-d0b708548748/go.mod h1:KEPIJ/ZdMFQ9tmmfv/uT9TjQEuEcZupCZBabuRGEC1k=
 github.com/livekit/psrpc v0.7.1 h1:ms37az0QTD3UXIWuUC5D/SkmKOlRMVRsI261eBWu/Vw=
 github.com/livekit/psrpc v0.7.1/go.mod h1:bZ4iHFQptTkbPnB0LasvRNu/OBYXEu1NA6O5BMFo9kk=
 github.com/livekit/server-sdk-go/v2 v2.16.2 h1:eQe24cka3X+5zUivezyL72nwtAJTWFXgibeiyJ/Jm+Y=

pkg/console/pipeline.go

@@ -72,6 +72,7 @@ type AudioPipeline struct {
 	mu sync.Mutex
 	fftBands [NumFFTBands]float64
 	muted    bool
+	paused   bool    // true when audio I/O is paused (e.g. text mode); mic frames are not sent to the agent
 	level    float64 // capture level in dB
 	playing  bool    // true when outputting real audio (not silence)
 
@@ -267,6 +268,19 @@ func (p *AudioPipeline) Muted() bool {
 	return p.muted
 }
 
+// SetPaused stops mic frames from being sent to the agent and drops any
+// queued playback. The hardware streams keep running so the pipeline can
+// resume instantly. Used to keep audio off the wire in text mode.
+func (p *AudioPipeline) SetPaused(paused bool) {
+	p.mu.Lock()
+	p.paused = paused
+	p.mu.Unlock()
+
+	if paused && p.playbackRing != nil {
+		p.playbackRing.Reset()
+	}
+}
+
 func (p *AudioPipeline) Level() float64 {
 	p.mu.Lock()
 	defer p.mu.Unlock()
@@ -389,6 +403,13 @@ func (p *AudioPipeline) speakerLoop(ctx context.Context) {
 			}
 		}
 
+		p.mu.Lock()
+		paused := p.paused
+		p.mu.Unlock()
+		if paused {
+			continue
+		}
+
 		p.computeMetrics(captureBuf)
 
 		_ = p.writeMessage(&agent.AgentSessionMessage{