J.A.R.V.I.S-rust/crates/jarvis-core/src/tts/piper.rs

//! Piper neural TTS backend (rhasspy/piper).
//!
//! Discovery order for binary:
//!   1. env `JARVIS_TTS_PIPER_BIN`
//!   2. `<exe_dir>/tools/piper/piper.exe`
//!   3. `<APP_DIR>/tools/piper/piper.exe`
//!
//! Discovery order for voice:
//!   1. env `JARVIS_TTS_PIPER_VOICE` (absolute path to .onnx)
//!   2. `<piper_dir>/voices/ru_RU-irina-medium.onnx`
//!   3. first `*.onnx` found in `<piper_dir>/voices/`

use std::path::{Path, PathBuf};

use super::{SpeakOpts, TtsBackend};
use crate::APP_DIR;

pub struct PiperBackend {
    binary: PathBuf,
    voice: PathBuf,
}

impl PiperBackend {
    pub fn try_new() -> Result<Self, String> {
        let binary = find_binary().ok_or_else(|| "piper.exe not found".to_string())?;
        let voice_dir = binary
            .parent()
            .map(|p| p.join("voices"))
            .unwrap_or_else(|| PathBuf::from("voices"));
        let voice = find_voice(&voice_dir).ok_or_else(|| {
            format!("no .onnx voice in {}", voice_dir.display())
        })?;
        Ok(Self { binary, voice })
    }

    pub fn binary_display(&self) -> String {
        self.binary.display().to_string()
    }

    pub fn voice_display(&self) -> String {
        self.voice
            .file_name()
            .map(|s| s.to_string_lossy().to_string())
            .unwrap_or_else(|| self.voice.display().to_string())
    }
}

impl TtsBackend for PiperBackend {
    fn name(&self) -> &'static str { "piper" }

    fn speak(&self, text: &str, opts: &SpeakOpts) {
        if text.trim().is_empty() {
            return;
        }
        let binary = self.binary.clone();
        let voice = self.voice.clone();
        let text = text.to_string();
        let detached = opts.detached;

        let run = move || synth_and_play(&binary, &voice, &text);

        if detached {
            std::thread::spawn(run);
        } else {
            run();
        }
    }
}

fn synth_and_play(binary: &Path, voice: &Path, text: &str) {
    let tmp = match tempfile::Builder::new()
        .prefix("jarvis-tts-")
        .suffix(".wav")
        .tempfile()
    {
        Ok(f) => f,
        Err(e) => {
            log::warn!("Piper: cannot create temp wav: {}", e);
            return;
        }
    };
    let wav_path = tmp.path().to_path_buf();
    drop(tmp);

    let mut child = match std::process::Command::new(binary)
        .arg("--model").arg(voice)
        .arg("--output_file").arg(&wav_path)
        .stdin(std::process::Stdio::piped())
        .stdout(std::process::Stdio::null())
        .stderr(std::process::Stdio::piped())
        .spawn()
    {
        Ok(c) => c,
        Err(e) => {
            log::warn!("Piper spawn failed: {}", e);
            return;
        }
    };

    if let Some(mut stdin) = child.stdin.take() {
        use std::io::Write;
        if let Err(e) = stdin.write_all(text.as_bytes()) {
            log::warn!("Piper stdin write failed: {}", e);
        }
    }

    match child.wait() {
        Ok(status) if status.success() => {
            play_wav(&wav_path);
        }
        Ok(status) => log::warn!("Piper exited with status {}", status),
        Err(e) => log::warn!("Piper wait failed: {}", e),
    }

    let _ = std::fs::remove_file(&wav_path);
}

#[cfg(target_os = "windows")]
fn play_wav(path: &Path) {
    let escaped = path.display().to_string().replace('\'', "''");
    let ps = format!(
        "$p = New-Object System.Media.SoundPlayer '{}'; $p.PlaySync()",
        escaped
    );
    let _ = std::process::Command::new("powershell")
        .args(["-NoProfile", "-ExecutionPolicy", "Bypass", "-Command", &ps])
        .stdout(std::process::Stdio::null())
        .stderr(std::process::Stdio::null())
        .status();
}

#[cfg(not(target_os = "windows"))]
fn play_wav(path: &Path) {
    log::info!("[Piper non-Windows stub] would play: {}", path.display());
}

fn find_binary() -> Option<PathBuf> {
    if let Ok(p) = std::env::var("JARVIS_TTS_PIPER_BIN") {
        let candidate = PathBuf::from(p);
        if candidate.is_file() {
            return Some(candidate);
        }
    }

    let bin_name = if cfg!(target_os = "windows") { "piper.exe" } else { "piper" };

    if let Ok(exe) = std::env::current_exe() {
        if let Some(dir) = exe.parent() {
            let c = dir.join("tools").join("piper").join(bin_name);
            if c.is_file() {
                return Some(c);
            }
        }
    }

    let c = APP_DIR.join("tools").join("piper").join(bin_name);
    if c.is_file() {
        return Some(c);
    }

    None
}

fn find_voice(voice_dir: &Path) -> Option<PathBuf> {
    if let Ok(p) = std::env::var("JARVIS_TTS_PIPER_VOICE") {
        let candidate = PathBuf::from(p);
        if candidate.is_file() {
            return Some(candidate);
        }
    }

    let preferred = voice_dir.join("ru_RU-irina-medium.onnx");
    if preferred.is_file() {
        return Some(preferred);
    }

    let entries = std::fs::read_dir(voice_dir).ok()?;
    for entry in entries.flatten() {
        let p = entry.path();
        if p.extension().and_then(|s| s.to_str()) == Some("onnx") {
            return Some(p);
        }
    }
    None
}