moltis-org · penso · Feb 21, 2026 · Feb 21, 2026 · Feb 21, 2026
diff --git a/Cargo.toml b/Cargo.toml
@@ -125,7 +125,7 @@ dashmap     = "6"
 hostname    = "0.4"
 http        = "1"
 hyper       = { features = ["http1", "http2", "server"], version = "1" }
-hyper-util  = { features = ["tokio", "server-auto", "http1", "http2", "service"], version = "0.1" }
+hyper-util  = { features = ["http1", "http2", "server-auto", "service", "tokio"], version = "0.1" }
 tower       = "0.5"
 tower-http  = "0.6"
 url         = "2"
@@ -185,7 +185,7 @@ moltis-tools      = { path = "crates/tools" }
 moltis-voice      = { path = "crates/voice" }
 
 [profile.release]
-lto            = "thin"
-strip          = true
-panic          = "abort"
-codegen-units  = 1
+codegen-units = 1
+lto           = "thin"
+panic         = "abort"
+strip         = true
diff --git a/crates/cli/Cargo.toml b/crates/cli/Cargo.toml
@@ -79,6 +79,7 @@ default = [
   "tailscale",
   "tls",
   "voice",
+  "voicebox",
   "web-ui",
 ]
 # Minimal build for memory-constrained devices (Raspberry Pi, etc.).
@@ -94,6 +95,7 @@ qmd                = ["moltis-gateway/qmd"]
 tailscale          = ["moltis-gateway/tailscale"]
 tls                = ["moltis-gateway/tls"]
 voice              = ["moltis-gateway/voice"]
+voicebox           = ["moltis-gateway/voicebox"]
 web-ui             = ["moltis-gateway/web-ui"]
 
 [lints]

diff --git a/crates/config/src/schema.rs b/crates/config/src/schema.rs
@@ -223,7 +223,7 @@ pub struct VoiceConfig {
 pub struct VoiceTtsConfig {
     /// Enable TTS globally.
     pub enabled: bool,
-    /// Default provider: "elevenlabs", "openai", "google", "piper", "coqui".
+    /// Default provider: "elevenlabs", "openai", "google", "piper", "coqui", "voicebox".
     pub provider: String,
     /// Provider IDs to list in the UI. Empty means list all.
     pub providers: Vec<String>,
@@ -237,6 +237,8 @@ pub struct VoiceTtsConfig {
     pub piper: VoicePiperTtsConfig,
     /// Coqui TTS (local server) settings.
     pub coqui: VoiceCoquiTtsConfig,
+    /// Voicebox TTS (local voice-cloning server) settings.
+    pub voicebox: VoiceVoiceboxTtsConfig,
 }
 
 impl Default for VoiceTtsConfig {
@@ -250,6 +252,7 @@ impl Default for VoiceTtsConfig {
             google: VoiceGoogleTtsConfig::default(),
             piper: VoicePiperTtsConfig::default(),
             coqui: VoiceCoquiTtsConfig::default(),
+            voicebox: VoiceVoiceboxTtsConfig::default(),
         }
     }
 }
@@ -353,6 +356,31 @@ impl Default for VoiceCoquiTtsConfig {
     }
 }
 
+/// Voicebox TTS (local voice-cloning server) configuration.
+#[derive(Debug, Clone, Serialize, Deserialize)]
+#[serde(default)]
+pub struct VoiceVoiceboxTtsConfig {
+    /// Voicebox server endpoint (default: http://localhost:8000).
+    pub endpoint: String,
+    /// Voice profile ID (must be created in Voicebox UI first).
+    pub profile_id: Option<String>,
+    /// Model size (e.g., "1.7B").
+    pub model_size: Option<String>,
+    /// Language code (e.g., "en").
+    pub language: Option<String>,
+}
+
+impl Default for VoiceVoiceboxTtsConfig {
+    fn default() -> Self {
+        Self {
+            endpoint: "http://localhost:8000".into(),
+            profile_id: None,
+            model_size: None,
+            language: None,
+        }
+    }
+}
+
 /// Voice STT configuration for moltis.toml.
 #[derive(Debug, Clone, Serialize, Deserialize)]
 #[serde(default)]

diff --git a/crates/config/src/validate.rs b/crates/config/src/validate.rs
@@ -457,6 +457,15 @@ fn build_schema_map() -> KnownKeys {
                                 ("endpoint", Leaf),
                             ])),
                         ),
+                        (
+                            "voicebox",
+                            Struct(HashMap::from([
+                                ("endpoint", Leaf),
+                                ("profile_id", Leaf),
+                                ("model_size", Leaf),
+                                ("language", Leaf),
+                            ])),
+                        ),
                     ])),
                 ),
                 (
@@ -969,6 +978,7 @@ fn check_semantic_warnings(config: &MoltisConfig, diagnostics: &mut Vec<Diagnost
         "google-tts",
         "piper",
         "coqui",
+        "voicebox",
     ];
     for (idx, provider) in config.voice.tts.providers.iter().enumerate() {
         if !valid_voice_tts_providers.contains(&provider.as_str()) {

diff --git a/crates/cron/Cargo.toml b/crates/cron/Cargo.toml
@@ -10,8 +10,8 @@ chrono         = { workspace = true }
 chrono-tz      = { workspace = true }
 cron           = { workspace = true }
 moltis-agents  = { workspace = true }
-moltis-config  = { workspace = true }
 moltis-common  = { workspace = true }
+moltis-config  = { workspace = true }
 moltis-metrics = { optional = true, workspace = true }
 serde          = { workspace = true }
 serde_json     = { workspace = true }

diff --git a/crates/gateway/Cargo.toml b/crates/gateway/Cargo.toml
@@ -12,9 +12,6 @@ async-trait        = { workspace = true }
 axum               = { workspace = true }
 axum-extra         = { features = ["cookie"], workspace = true }
 axum-server        = { optional = true, workspace = true }
-hyper              = { optional = true, workspace = true }
-hyper-util         = { optional = true, workspace = true }
-tower              = { optional = true, workspace = true }
 base64             = { workspace = true }
 bytes              = { workspace = true }
 chrono             = { features = ["serde"], optional = true, workspace = true }
@@ -23,6 +20,8 @@ dashmap            = { workspace = true }
 futures            = { workspace = true }
 gix                = { workspace = true }
 hostname           = { workspace = true }
+hyper              = { optional = true, workspace = true }
+hyper-util         = { optional = true, workspace = true }
 include_dir        = { optional = true, workspace = true }
 moltis-agents      = { workspace = true }
 moltis-browser     = { workspace = true }
@@ -70,6 +69,7 @@ tokio-stream       = { workspace = true }
 tokio-tungstenite  = { workspace = true }
 tokio-util         = { workspace = true }
 toml               = { workspace = true }
+tower              = { optional = true, workspace = true }
 tower-http         = { features = ["catch-panic", "compression-gzip", "compression-zstd", "cors", "limit", "request-id", "sensitive-headers", "set-header", "trace"], workspace = true }
 tracing            = { workspace = true }
 tracing-subscriber = { workspace = true }
@@ -118,6 +118,7 @@ tls = [
   "dep:tower",
 ]
 voice = ["dep:moltis-voice"]
+voicebox = ["moltis-voice/voicebox", "voice"]
 web-ui = ["dep:askama", "dep:chrono", "dep:include_dir", "dep:portable-pty", "dep:which"]
 
 [dev-dependencies]

diff --git a/crates/gateway/src/voice.rs b/crates/gateway/src/voice.rs
@@ -25,6 +25,9 @@ use moltis_voice::{
     WhisperCliStt, WhisperStt, strip_ssml_tags,
 };
 
+#[cfg(feature = "voicebox")]
+use moltis_voice::{VoiceboxTts, VoiceboxTtsConfig};
+
 #[cfg(feature = "voice")]
 use crate::services::TtsService;
 
@@ -123,6 +126,13 @@ impl LiveTtsService {
                 speaker: None,
                 language: None,
             },
+            #[cfg(feature = "voicebox")]
+            voicebox: VoiceboxTtsConfig {
+                endpoint: cfg.voice.tts.voicebox.endpoint.clone(),
+                profile_id: cfg.voice.tts.voicebox.profile_id.clone(),
+                model_size: cfg.voice.tts.voicebox.model_size.clone(),
+                language: cfg.voice.tts.voicebox.language.clone(),
+            },
         }
     }
 
@@ -163,13 +173,23 @@ impl LiveTtsService {
                     None
                 }
             },
+            #[cfg(feature = "voicebox")]
+            TtsProviderId::Voicebox => {
+                let voicebox = VoiceboxTts::new(&config.voicebox);
+                if voicebox.is_configured() {
+                    Some(Box::new(voicebox) as Box<dyn TtsProvider + Send + Sync>)
+                } else {
+                    None
+                }
+            },
         }
     }
 
     /// List all providers with their configuration status.
     fn list_providers() -> Vec<(TtsProviderId, bool)> {
         let config = Self::load_config();
-        vec![
+        #[allow(unused_mut)]
+        let mut providers = vec![
             (
                 TtsProviderId::ElevenLabs,
                 config.elevenlabs.api_key.is_some(),
@@ -178,7 +198,10 @@ impl LiveTtsService {
             (TtsProviderId::Google, config.google.api_key.is_some()),
             (TtsProviderId::Piper, config.piper.model_path.is_some()),
             (TtsProviderId::Coqui, true), // Always available if server running
-        ]
+        ];
+        #[cfg(feature = "voicebox")]
+        providers.push((TtsProviderId::Voicebox, true)); // Always available if server running
+        providers
     }
 }
 
@@ -804,8 +827,13 @@ mod tests {
         let providers = service.providers().await.unwrap();
 
         let providers_arr = providers.as_array().unwrap();
-        // 5 providers: elevenlabs, openai, google, piper, coqui
-        assert_eq!(providers_arr.len(), 5);
+        // Base providers: elevenlabs, openai, google, piper, coqui
+        let expected_count = if cfg!(feature = "voicebox") {
+            6
+        } else {
+            5
+        };
+        assert_eq!(providers_arr.len(), expected_count);
 
         let ids: Vec<_> = providers_arr
             .iter()

diff --git a/crates/tools/src/web_search.rs b/crates/tools/src/web_search.rs
@@ -823,7 +823,7 @@ mod tests {
     async fn test_brave_missing_api_key_returns_hint() {
         let tool = brave_tool();
         let result = tool
-            .execute(serde_json::json!({"query": "test"}))
+            .search_brave("test", 5, &serde_json::json!({}), None, "")
             .await
             .unwrap();
         assert!(result["error"].as_str().unwrap().contains("not configured"));
@@ -834,7 +834,7 @@ mod tests {
     async fn test_perplexity_missing_api_key_returns_hint() {
         let tool = perplexity_tool();
         let result = tool
-            .execute(serde_json::json!({"query": "test"}))
+            .search_perplexity("test", "", "https://api.perplexity.ai", "sonar-pro")
             .await
             .unwrap();
         assert!(result["error"].as_str().unwrap().contains("not configured"));

diff --git a/crates/voice/Cargo.toml b/crates/voice/Cargo.toml
@@ -21,6 +21,9 @@ tracing     = { workspace = true }
 urlencoding = "2"
 which       = "7"
 
+[features]
+voicebox = []
+
 [dev-dependencies]
 tokio    = { features = ["macros", "test-util"], workspace = true }
 wiremock = "0.6"

diff --git a/crates/voice/src/config.rs b/crates/voice/src/config.rs
@@ -22,6 +22,9 @@ pub enum TtsProviderId {
     Piper,
     #[serde(rename = "coqui")]
     Coqui,
+    #[cfg(feature = "voicebox")]
+    #[serde(rename = "voicebox")]
+    Voicebox,
 }
 
 impl fmt::Display for TtsProviderId {
@@ -32,6 +35,8 @@ impl fmt::Display for TtsProviderId {
             Self::Google => write!(f, "google"),
             Self::Piper => write!(f, "piper"),
             Self::Coqui => write!(f, "coqui"),
+            #[cfg(feature = "voicebox")]
+            Self::Voicebox => write!(f, "voicebox"),
         }
     }
 }
@@ -45,6 +50,8 @@ impl TtsProviderId {
             "google" => Some(Self::Google),
             "piper" => Some(Self::Piper),
             "coqui" => Some(Self::Coqui),
+            #[cfg(feature = "voicebox")]
+            "voicebox" => Some(Self::Voicebox),
             _ => None,
         }
     }
@@ -57,6 +64,8 @@ impl TtsProviderId {
             Self::Google => "Google Cloud TTS",
             Self::Piper => "Piper",
             Self::Coqui => "Coqui TTS",
+            #[cfg(feature = "voicebox")]
+            Self::Voicebox => "Voicebox",
         }
     }
 
@@ -68,6 +77,8 @@ impl TtsProviderId {
             Self::Google,
             Self::Piper,
             Self::Coqui,
+            #[cfg(feature = "voicebox")]
+            Self::Voicebox,
         ]
     }
 }
@@ -200,6 +211,10 @@ pub struct TtsConfig {
 
     /// Coqui TTS (local) settings.
     pub coqui: CoquiTtsConfig,
+
+    /// Voicebox TTS (local voice-cloning) settings.
+    #[cfg(feature = "voicebox")]
+    pub voicebox: VoiceboxTtsConfig,
 }
 
 impl Default for TtsConfig {
@@ -214,6 +229,8 @@ impl Default for TtsConfig {
             google: GoogleTtsConfig::default(),
             piper: PiperTtsConfig::default(),
             coqui: CoquiTtsConfig::default(),
+            #[cfg(feature = "voicebox")]
+            voicebox: VoiceboxTtsConfig::default(),
         }
     }
 }
@@ -356,6 +373,36 @@ impl Default for CoquiTtsConfig {
     }
 }
 
+/// Voicebox TTS (local voice-cloning) configuration.
+#[cfg(feature = "voicebox")]
+#[derive(Debug, Clone, Serialize, Deserialize)]
+#[serde(default)]
+pub struct VoiceboxTtsConfig {
+    /// Voicebox server endpoint (default: http://localhost:8000).
+    pub endpoint: String,
+
+    /// Voice profile ID (must be created in Voicebox UI first).
+    pub profile_id: Option<String>,
+
+    /// Model size (e.g., "1.7B").
+    pub model_size: Option<String>,
+
+    /// Language code (e.g., "en").
+    pub language: Option<String>,
+}
+
+#[cfg(feature = "voicebox")]
+impl Default for VoiceboxTtsConfig {
+    fn default() -> Self {
+        Self {
+            endpoint: "http://localhost:8000".into(),
+            profile_id: None,
+            model_size: None,
+            language: None,
+        }
+    }
+}
+
 /// ElevenLabs Scribe STT configuration.
 #[derive(Debug, Clone, Default, Serialize, Deserialize)]
 #[serde(default)]
@@ -662,6 +709,8 @@ mod tests {
                 google: GoogleTtsConfig::default(),
                 piper: PiperTtsConfig::default(),
                 coqui: CoquiTtsConfig::default(),
+                #[cfg(feature = "voicebox")]
+                voicebox: VoiceboxTtsConfig::default(),
             },
             stt: SttConfig::default(),
         };

diff --git a/crates/voice/src/lib.rs b/crates/voice/src/lib.rs
@@ -24,3 +24,6 @@ pub use {
         strip_ssml_tags,
     },
 };
+
+#[cfg(feature = "voicebox")]
+pub use {config::VoiceboxTtsConfig, tts::VoiceboxTts};
diff --git a/crates/voice/src/tts/mod.rs b/crates/voice/src/tts/mod.rs
@@ -5,7 +5,11 @@ mod elevenlabs;
 mod google;
 mod openai;
 mod piper;
+#[cfg(feature = "voicebox")]
+mod voicebox;
 
+#[cfg(feature = "voicebox")]
+pub use voicebox::VoiceboxTts;
 pub use {
     coqui::CoquiTts, elevenlabs::ElevenLabsTts, google::GoogleTts, openai::OpenAiTts,
     piper::PiperTts,