From 70cb975f225f3ff18b10534e8342cd62e0bf66ce Mon Sep 17 00:00:00 2001
From: soryu <soryu@soryu.co>
Date: Thu, 29 Jan 2026 03:02:51 +0000
Subject: Fix model loading for TTS / speak

---
 makima/src/bin/makima.rs        |  1 +
 makima/src/daemon/cli/server.rs |  4 ++++
 makima/src/server/state.rs      | 12 ++++++++++--
 3 files changed, 15 insertions(+), 2 deletions(-)

(limited to 'makima/src')

diff --git a/makima/src/bin/makima.rs b/makima/src/bin/makima.rs
index 8e83565..c637c36 100644
--- a/makima/src/bin/makima.rs
+++ b/makima/src/bin/makima.rs
@@ -49,6 +49,7 @@ async fn run_server(
         &args.parakeet_model_dir,
         &args.parakeet_eou_dir,
         &args.sortformer_model_path,
+        &args.qwen3_tts_dir,
     );
 
     // Connect to database if URL provided
diff --git a/makima/src/daemon/cli/server.rs b/makima/src/daemon/cli/server.rs
index 371a912..81dafc9 100644
--- a/makima/src/daemon/cli/server.rs
+++ b/makima/src/daemon/cli/server.rs
@@ -33,6 +33,10 @@ pub struct ServerArgs {
     )]
     pub sortformer_model_path: String,
 
+    /// Path to Qwen3-TTS model directory
+    #[arg(long, env = "QWEN3_TTS_DIR", default_value = "models/qwen3-tts")]
+    pub qwen3_tts_dir: String,
+
     /// PostgreSQL connection URI
     #[arg(long, env = "POSTGRES_CONNECTION_URI")]
     pub database_url: Option<String>,
diff --git a/makima/src/server/state.rs b/makima/src/server/state.rs
index 041b101..4650a57 100644
--- a/makima/src/server/state.rs
+++ b/makima/src/server/state.rs
@@ -557,6 +557,7 @@ pub struct ModelConfig {
     pub parakeet_model_dir: String,
     pub parakeet_eou_dir: String,
     pub sortformer_model_path: String,
+    pub qwen3_tts_dir: String,
 }
 
 /// Lazily-loaded ML models.
@@ -615,10 +616,12 @@ impl AppState {
     /// * `parakeet_model_dir` - Path to the Parakeet TDT model directory
     /// * `parakeet_eou_dir` - Path to the Parakeet EOU model directory
     /// * `sortformer_model_path` - Path to the Sortformer diarization model file
+    /// * `qwen3_tts_dir` - Path to the Qwen3-TTS model directory
     pub fn new(
         parakeet_model_dir: &str,
         parakeet_eou_dir: &str,
         sortformer_model_path: &str,
+        qwen3_tts_dir: &str,
     ) -> Self {
         // Create broadcast channels with buffer for 256 messages
         let (file_updates, _) = broadcast::channel(256);
@@ -662,6 +665,7 @@ impl AppState {
                 parakeet_model_dir: parakeet_model_dir.to_string(),
                 parakeet_eou_dir: parakeet_eou_dir.to_string(),
                 sortformer_model_path: sortformer_model_path.to_string(),
+                qwen3_tts_dir: qwen3_tts_dir.to_string(),
             }),
             ml_models: OnceCell::new(),
             db_pool: None,
@@ -687,11 +691,15 @@ impl AppState {
     /// The TTS engine is loaded on first Speak connection using the Qwen3 backend.
     /// Returns a reference to the engine, or an error if loading fails.
     pub async fn get_tts_engine(&self) -> Result<&dyn TtsEngine, Box<dyn std::error::Error + Send + Sync>> {
+        let tts_dir = self.model_config.as_ref().map(|c| c.qwen3_tts_dir.as_str());
         self.tts_engine.get_or_try_init(|| async {
-            tracing::info!("Lazy-loading TTS engine (Qwen3) on first Speak connection...");
+            tracing::info!(
+                model_dir = ?tts_dir,
+                "Lazy-loading TTS engine (Qwen3) on first Speak connection..."
+            );
             let engine = crate::tts::TtsEngineFactory::create(
                 crate::tts::TtsBackend::Qwen3,
-                None, // Use default model directory
+                tts_dir,
             ).map_err(|e| -> Box<dyn std::error::Error + Send + Sync> {
                 Box::new(e)
             })?;
-- 
cgit v1.2.3