feat: piper now not a subprocess, cache system for TTS

1 files changed, 17 insertions, 17 deletions

diff --git a/src/events/mod.rs b/src/events/mod.rs
index e4f1b22..87333c8 100644
--- a/src/events/mod.rs
+++ b/src/events/mod.rs
@@ -28,15 +28,14 @@ fn format_duration(seconds: f64) -> String {
 }
 
 async fn play_tts_in_channel(
-    config: crate::config::Config,
-    http_client: reqwest::Client,
+    tts: std::sync::Arc<crate::tts::TtsState>,
     voice_manager: crate::voice::VoiceManager,
     guild_id: serenity::GuildId,
     channel_id: serenity::ChannelId,
     text: String,
 ) {
-    let audio_data = match crate::tts::synthesize(&config, &http_client, &text).await {
-        Ok((bytes, _)) => bytes,
+    let audio_data = match tts.synthesize(&text).await {
+        Ok(bytes) => bytes,
         Err(e) => {
             warn!("Failed to synthesize TTS: {e}");
             return;
@@ -51,13 +50,7 @@ async fn play_tts_in_channel(
         }
     };
 
-    let input = match crate::voice::decode_audio(&audio_data) {
-        Ok(i) => i,
-        Err(e) => {
-            warn!("Failed to decode TTS audio: {e}");
-            return;
-        }
-    };
+    let input = crate::tts::pcm_bytes_to_input(audio_data);
 
     let mut call_lock = call.lock().await;
     if call_lock.current_channel().is_some() {
@@ -97,6 +90,16 @@ pub async fn handle_voice_state_update(
         .map(|m| m.display_name().to_string())
         .unwrap_or_default();
 
+    // Pre-generate all 6 event phrases for this user (background, non-blocking).
+    // First event per user pays ~120ms; subsequent events are instant from cache.
+    {
+        let tts = data.tts.clone();
+        let name = display_name.clone();
+        tokio::spawn(async move {
+            tts.pre_generate(&name).await;
+        });
+    }
+
     let now = current_ts as f64;
 
     let user_id_str = user_id.to_string();
@@ -293,8 +296,7 @@ pub async fn handle_voice_state_update(
             && let Some(cid) = tts_channel.or(new_channel)
         {
             tokio::spawn(play_tts_in_channel(
-                data.config.clone(),
-                data.http_client.clone(),
+                data.tts.clone(),
                 data.voice_manager.clone(),
                 gid,
                 cid,
@@ -314,8 +316,7 @@ pub async fn handle_voice_state_update(
             && let Some(cid) = tts_channel.or(new_channel)
         {
             tokio::spawn(play_tts_in_channel(
-                data.config.clone(),
-                data.http_client.clone(),
+                data.tts.clone(),
                 data.voice_manager.clone(),
                 gid,
                 cid,
@@ -330,8 +331,7 @@ pub async fn handle_voice_state_update(
         && let Some(gid) = guild_id
     {
         tokio::spawn(play_tts_in_channel(
-            data.config.clone(),
-            data.http_client.clone(),
+            data.tts.clone(),
             data.voice_manager.clone(),
             gid,
             cid,