Polish Huddle voice controls

desktop/src-tauri/src/huddle/commands.rs

@@ -0,0 +1,132 @@
+//! Small Huddle controls that mutate an active session.
+
+use std::sync::{atomic::Ordering, Arc};
+
+use tauri::State;
+use uuid::Uuid;
+
+use crate::{app_state::AppState, events, relay::submit_event};
+
+use super::{relay_api::validate_pubkey_hex, HuddlePhase};
+
+/// Update the clickable microphone control independently from the PTT shortcut.
+#[tauri::command]
+pub fn set_huddle_manual_mic_unmuted(
+    enabled: bool,
+    state: State<'_, AppState>,
+) -> Result<(), String> {
+    let huddle = state.huddle()?;
+    if !matches!(huddle.phase, HuddlePhase::Connected | HuddlePhase::Active) {
+        return Err("no active huddle".to_string());
+    }
+    huddle.manual_mic_unmuted.store(enabled, Ordering::Release);
+    Ok(())
+}
+
+/// Immediately interrupt the agent utterance that is currently speaking.
+#[tauri::command]
+pub fn interrupt_huddle_speech(
+    agent_pubkey: String,
+    state: State<'_, AppState>,
+) -> Result<(), String> {
+    validate_pubkey_hex(&agent_pubkey)?;
+    let tts_pipeline = {
+        let huddle = state.huddle()?;
+        if !matches!(huddle.phase, HuddlePhase::Connected | HuddlePhase::Active) {
+            return Err("no active huddle".to_string());
+        }
+        huddle.tts_pipeline.as_ref().map(Arc::clone)
+    };
+    if let Some(tts_pipeline) = tts_pipeline {
+        tts_pipeline.cancel_active_speaker(&agent_pubkey);
+    }
+    Ok(())
+}
+
+/// Remove an agent from the active huddle without removing its parent-channel
+/// membership. Keeping the parent membership intact means it remains available
+/// to rejoin this huddle from the agent picker.
+#[tauri::command]
+pub async fn remove_agent_from_huddle(
+    agent_pubkey: String,
+    state: State<'_, AppState>,
+) -> Result<(), String> {
+    validate_pubkey_hex(&agent_pubkey)?;
+
+    let (ephemeral_channel_id, huddle_generation) = {
+        let huddle = state.huddle()?;
+        if !matches!(huddle.phase, HuddlePhase::Connected | HuddlePhase::Active) {
+            return Err("no active huddle".to_string());
+        }
+
+        let is_huddle_agent = huddle
+            .agent_pubkeys
+            .lock()
+            .unwrap_or_else(|error| error.into_inner())
+            .iter()
+            .any(|pubkey| pubkey.eq_ignore_ascii_case(&agent_pubkey));
+        if !is_huddle_agent {
+            return Err("agent is not in this huddle".to_string());
+        }
+
+        (
+            huddle
+                .ephemeral_channel_id
+                .clone()
+                .ok_or("no ephemeral channel")?,
+            huddle.huddle_generation,
+        )
+    };
+
+    let ephemeral_channel_uuid =
+        Uuid::parse_str(&ephemeral_channel_id).map_err(|error| error.to_string())?;
+    submit_event(
+        events::build_remove_member(ephemeral_channel_uuid, &agent_pubkey)?,
+        &state,
+    )
+    .await?;
+
+    let (roster_changed, tts_pipeline) = {
+        let mut huddle = state.huddle()?;
+        if !huddle.is_current_huddle(&ephemeral_channel_id, huddle_generation) {
+            (false, None)
+        } else {
+            let mut agent_pubkeys = huddle
+                .agent_pubkeys
+                .lock()
+                .unwrap_or_else(|error| error.into_inner());
+            let initial_count = agent_pubkeys.len();
+            agent_pubkeys.retain(|pubkey| !pubkey.eq_ignore_ascii_case(&agent_pubkey));
+            let changed = agent_pubkeys.len() != initial_count;
+            drop(agent_pubkeys);
+
+            if changed {
+                huddle
+                    .participants
+                    .retain(|pubkey| !pubkey.eq_ignore_ascii_case(&agent_pubkey));
+                if let Some(settings_pubkey) = huddle
+                    .agent_voice_settings
+                    .keys()
+                    .find(|pubkey| pubkey.eq_ignore_ascii_case(&agent_pubkey))
+                    .cloned()
+                {
+                    huddle.agent_voice_settings.remove(&settings_pubkey);
+                }
+            }
+            let tts_pipeline = changed
+                .then_some(huddle.tts_pipeline.as_ref())
+                .flatten()
+                .map(Arc::clone);
+            (changed, tts_pipeline)
+        }
+    };
+
+    if let Some(tts_pipeline) = tts_pipeline {
+        tts_pipeline.cancel_speaker(&agent_pubkey);
+    }
+    if roster_changed {
+        state.emit_huddle_state_changed();
+    }
+
+    Ok(())
+}

desktop/src-tauri/src/huddle/mod.rs

@@ -27,6 +27,7 @@ mod agent_tts_routing;
 pub mod agent_voice;
 pub mod agents;
 pub mod audio_output;
+mod commands;
 pub mod jitter;
 pub mod models;
 pub mod pipeline;
@@ -67,6 +68,9 @@ pub(super) fn drain_until_shutdown<T>(
 
 // ── Re-exports ────────────────────────────────────────────────────────────────
 
+pub use commands::{
+    interrupt_huddle_speech, remove_agent_from_huddle, set_huddle_manual_mic_unmuted,
+};
 pub use state::{HuddleJoinInfo, HuddlePhase, HuddleState, VoiceInputMode};
 pub use transcription::{set_huddle_transcription_enabled, start_stt_pipeline};
 pub use tts_settings::set_tts_enabled;
@@ -868,19 +872,41 @@ pub async fn speak_agent_message(
 
     let sender = {
         let hs = state.huddle()?;
+        let agent_is_present = hs
+            .agent_pubkeys
+            .lock()
+            .unwrap_or_else(|error| error.into_inner())
+            .iter()
+            .any(|pubkey| pubkey.eq_ignore_ascii_case(&speaker_pubkey));
+        if !agent_is_present {
+            eprintln!(
+                "buzz-desktop: tts stage=queue status=dropped reason=speaker_removed route_id={route_id}"
+            );
+            return Ok(());
+        }
         hs.tts_pipeline
             .as_ref()
             .map(|pipeline| pipeline.text_sender())
+            .map(|sender| {
+                let speaker_generation = sender.speaker_generation(&speaker_pubkey);
+                (sender, speaker_generation)
+            })
     };
-    let Some(sender) = sender else {
+    let Some((sender, speaker_generation)) = sender else {
         eprintln!(
             "buzz-desktop: tts stage=invoke status=failed reason=unavailable route_id={route_id}"
         );
         return Err("Agent text to speech is enabled but its audio pipeline is unavailable".into());
     };
     enqueue_agent_tts_text(route_id, text, move |route_id, text| {
         sender
-            .send(route_id, speaker_pubkey, voice_reference, text)
+            .send(
+                route_id,
+                speaker_pubkey,
+                speaker_generation,
+                voice_reference,
+                text,
+            )
             .map_err(|error| format!("TTS queue closed while waiting to enqueue: {error}"))
     })
     .await

desktop/src-tauri/src/huddle/pipeline.rs

@@ -315,6 +315,7 @@ pub(crate) async fn maybe_start_stt_pipeline(
         expected_generation,
         stt_starting,
         ptt_active_for_stt,
+        manual_mic_unmuted_for_stt,
         old_stt,
     ) = {
         let mut hs = state.huddle()?;
@@ -338,21 +339,32 @@ pub(crate) async fn maybe_start_stt_pipeline(
         } else {
             None
         };
+        let manual_mic_unmuted = if hs.voice_input_mode == VoiceInputMode::PushToTalk {
+            Some(Arc::clone(&hs.manual_mic_unmuted))
+        } else {
+            None
+        };
         (
             Arc::clone(&hs.tts_active),
             Arc::clone(&hs.agent_pubkeys),
             Arc::clone(&hs.session_generation),
             hs.session_generation.load(Ordering::Acquire),
             stt_starting,
             ptt,
+            manual_mic_unmuted,
             old,
         )
     };
     // Drop the old pipeline OUTSIDE the lock — thread join happens here.
     drop(old_stt);
 
     let constructed = tokio::task::spawn_blocking(move || {
-        stt::SttPipeline::new(model_dir, tts_active, ptt_active_for_stt)
+        stt::SttPipeline::new(
+            model_dir,
+            tts_active,
+            ptt_active_for_stt,
+            manual_mic_unmuted_for_stt,
+        )
     })
     .await;
     let (pipeline, text_rx) = match constructed {

desktop/src-tauri/src/huddle/state.rs

@@ -15,7 +15,7 @@ use super::{stt, tts};
 
 /// Voice input mode: push-to-talk (PTT) or voice-activity detection (VAD).
 ///
-/// PTT: mic is gated by a global shortcut (Ctrl+Space). Pressing the key sets
+/// PTT (the default): mic is gated by a global shortcut (Ctrl+Space). Pressing the key sets
 /// `ptt_active` and immediately cancels any playing TTS. Releasing the key
 /// (after a 200 ms delay) stops mic capture and flushes the utterance.
 ///
@@ -26,8 +26,8 @@ use super::{stt, tts};
 #[derive(Debug, Clone, Default, Serialize, Deserialize, PartialEq)]
 #[serde(rename_all = "snake_case")]
 pub enum VoiceInputMode {
-    PushToTalk,
     #[default]
+    PushToTalk,
     VoiceActivity,
 }
 
@@ -135,6 +135,10 @@ pub struct HuddleState {
     /// Shared with the STT pipeline for mic gating.
     #[serde(skip)]
     pub ptt_active: Arc<AtomicBool>,
+    /// True while the clickable microphone control is manually unmuted.
+    /// In PTT mode, either this flag or `ptt_active` opens the STT gate.
+    #[serde(skip)]
+    pub manual_mic_unmuted: Arc<AtomicBool>,
 }
 
 fn serialize_agent_pubkeys<S>(v: &Arc<Mutex<Vec<String>>>, s: S) -> Result<S::Ok, S::Error>
@@ -190,6 +194,7 @@ impl Clone for HuddleState {
             session_generation: Arc::clone(&self.session_generation),
             voice_input_mode: self.voice_input_mode.clone(),
             ptt_active: Arc::clone(&self.ptt_active),
+            manual_mic_unmuted: Arc::clone(&self.manual_mic_unmuted),
         }
     }
 }
@@ -221,6 +226,7 @@ impl Default for HuddleState {
             session_generation: Arc::new(AtomicU64::new(0)),
             voice_input_mode: VoiceInputMode::default(),
             ptt_active: Arc::new(AtomicBool::new(false)),
+            manual_mic_unmuted: Arc::new(AtomicBool::new(true)),
         }
     }
 }
@@ -332,6 +338,13 @@ mod tests {
         assert!(!state.maybe_auto_enable_transcription_for_agents());
     }
 
+    #[test]
+    fn defaults_to_push_to_talk_with_an_open_microphone() {
+        let state = HuddleState::default();
+        assert_eq!(state.voice_input_mode, super::VoiceInputMode::PushToTalk);
+        assert!(state.manual_mic_unmuted.load(Ordering::Acquire));
+    }
+
     #[test]
     fn explicit_user_disable_is_not_undone_by_agent_presence() {
         let mut state = HuddleState::default();