wip

crates/speech/Cargo.toml

@@ -11,6 +11,7 @@ gpui.workspace = true
 gpui-kit = { workspace = true, features = ["speech"] }
 flume.workspace = true
 instant.workspace = true
+tts.workspace = true
 
 [target.'cfg(not(target_arch = "wasm32"))'.dependencies]
 common = { path = "../common" }

crates/speech/src/lib.rs

@@ -2,6 +2,7 @@
 pub mod model;
 #[cfg(not(target_arch = "wasm32"))]
 mod parakeet;
+pub mod speaker;
 #[cfg(target_arch = "wasm32")]
 mod web;
 

crates/speech/src/speaker.rs

@@ -0,0 +1,47 @@
+use tts::Tts;
+
+/// Reads text aloud through the platform's speech backend, created on first use.
+pub struct Speaker {
+    tts: Option<Tts>,
+}
+
+impl Default for Speaker {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl Speaker {
+    pub fn new() -> Self {
+        Self { tts: None }
+    }
+
+    /// Speak the text, interrupting what this speaker is already saying.
+    pub fn speak(&mut self, text: &str) -> Result<(), tts::Error> {
+        let tts = match self.tts.as_ref() {
+            Some(tts) => tts,
+            None => self.tts.insert(Tts::default()?),
+        };
+
+        tts.speak(text, true)?;
+        Ok(())
+    }
+
+    /// Stop speaking; a speaker that never spoke is already silent.
+    pub fn stop(&mut self) -> Result<(), tts::Error> {
+        match self.tts.as_ref() {
+            Some(tts) => tts.stop()?,
+            None => return Ok(()),
+        }
+
+        Ok(())
+    }
+
+    /// Best-effort speaking state; a speaker that never spoke is silent.
+    pub fn speaking(&self) -> bool {
+        self.tts
+            .as_ref()
+            .and_then(|tts| tts.is_speaking().ok())
+            .unwrap_or(false)
+    }
+}