lesavka/server/src/audio/voice_input.rs

pub struct ClipTap {
    buf: Vec<u8>,
    tag: &'static str,
    next_dump: Instant,
    period: Duration,
}

impl ClipTap {
    pub fn new(tag: &'static str, period: Duration) -> Self {
        Self {
            buf: Vec::with_capacity(260_000),
            tag,
            next_dump: Instant::now() + period,
            period,
        }
    }
    pub fn feed(&mut self, bytes: &[u8]) {
        self.buf.extend_from_slice(bytes);
        if self.buf.len() > 256_000 {
            self.buf.drain(..self.buf.len() - 256_000);
        }
        if Instant::now() >= self.next_dump {
            self.flush();
            self.next_dump += self.period;
        }
    }
    pub fn flush(&mut self) {
        if self.buf.is_empty() {
            return;
        }
        let ts = chrono::Local::now().format("%Y%m%d-%H%M%S");
        let path = format!("/tmp/{}-{}.aac", self.tag, ts);
        let _ = std::fs::write(&path, &self.buf);
        self.buf.clear();
    }
}
impl Drop for ClipTap {
    fn drop(&mut self) {
        self.flush()
    }
}

// ────────────────────── microphone sink  ────────────────────────────────
pub struct Voice {
    appsrc: gst_app::AppSrc,
    _pipe: gst::Pipeline, // keep pipeline alive
    tap: ClipTap,
}

impl Drop for Voice {
    fn drop(&mut self) {
        let _ = self._pipe.set_state(gst::State::Null);
    }
}

fn voice_input_caps() -> gst::Caps {
    gst::Caps::builder("audio/mpeg")
        .field("mpegversion", 4i32)
        .field("stream-format", "adts")
        .field("rate", 48_000i32)
        .field("channels", 2i32)
        .build()
}

impl Voice {
    #[cfg(coverage)]
    pub async fn new(_alsa_dev: &str) -> anyhow::Result<Self> {
        gst::init().context("gst init")?;

        let pipeline = gst::Pipeline::new();
        let appsrc = gst::ElementFactory::make("appsrc")
            .build()
            .context("make appsrc")?
            .downcast::<gst_app::AppSrc>()
            .expect("appsrc");
        appsrc.set_format(gst::Format::Time);
        appsrc.set_is_live(true);
        appsrc.set_caps(Some(&voice_input_caps()));

        let sink = gst::ElementFactory::make("fakesink")
            .build()
            .context("make fakesink")?;
        pipeline.add_many(&[appsrc.upcast_ref(), &sink])?;
        gst::Element::link_many(&[appsrc.upcast_ref(), &sink])?;
        start_pipeline_or_reset(&pipeline, "starting voice pipeline")?;

        Ok(Self {
            appsrc,
            _pipe: pipeline,
            tap: ClipTap::new("voice", Duration::from_secs(60)),
        })
    }

    #[cfg(not(coverage))]
    pub async fn new(alsa_dev: &str) -> anyhow::Result<Self> {
        use gst::prelude::*;

        gst::init().context("gst init")?;

        // pipeline
        let pipeline = gst::Pipeline::new();

        // elements
        let appsrc = gst::ElementFactory::make("appsrc")
            .build()
            .context("make appsrc")?
            .downcast::<gst_app::AppSrc>()
            .unwrap();

        // dedicated AppSrc helpers exist and avoid the needless `?`
        appsrc.set_caps(Some(&voice_input_caps()));
        appsrc.set_format(gst::Format::Time);
        appsrc.set_is_live(true);

        let decodebin = gst::ElementFactory::make("decodebin")
            .build()
            .context("make decodebin")?;
        let convert = gst::ElementFactory::make("audioconvert")
            .build()
            .context("make audioconvert")?;
        let resample = gst::ElementFactory::make("audioresample")
            .build()
            .context("make audioresample")?;
        let caps = gst::Caps::builder("audio/x-raw")
            .field("format", "S16LE")
            .field("channels", 2i32)
            .field("rate", 48_000i32)
            .build();
        let capsfilter = gst::ElementFactory::make("capsfilter")
            .property("caps", &caps)
            .build()
            .context("make capsfilter")?;
        let alsa_sink = gst::ElementFactory::make("alsasink")
            .build()
            .context("make alsasink")?;

        alsa_sink.set_property("device", alsa_dev);
        alsa_sink.set_property("sync", false);
        alsa_sink.set_property("async", false);
        alsa_sink.set_property("enable-last-sample", false);

        pipeline.add_many([
            appsrc.upcast_ref(),
            &decodebin,
            &convert,
            &resample,
            &capsfilter,
            &alsa_sink,
        ])?;
        appsrc.link(&decodebin)?;
        gst::Element::link_many([&convert, &resample, &capsfilter, &alsa_sink])?;

        /*------------ decodebin autolink ----------------*/
        let convert_sink = convert
            .static_pad("sink")
            .context("audioconvert sink pad")?;
        decodebin.connect_pad_added(move |_db, pad| {
            if convert_sink.is_linked() {
                return;
            }
            let caps = pad.current_caps().unwrap_or_else(|| pad.query_caps(None));
            let is_audio = caps
                .structure(0)
                .map(|s| s.name().starts_with("audio/"))
                .unwrap_or(false);
            if !is_audio {
                return;
            }
            let _ = pad.link(&convert_sink);
        });

        let bus = pipeline.bus().context("voice pipeline bus")?;

        // underrun ≠ error – just show a warning
        // let _id = alsa_sink.connect("underrun", false, |_| {
        //     tracing::warn!("⚠️  USB playback underrun – host muted/not reading");
        //     None
        // });

        start_pipeline_or_reset(&pipeline, "starting voice pipeline")?;
        spawn_pipeline_bus_logger(bus, "voice", "🎤 voice pipeline ▶️");

        Ok(Self {
            appsrc,
            _pipe: pipeline,
            tap: ClipTap::new("voice", Duration::from_secs(60)),
        })
    }

    pub fn push(&mut self, pkt: &AudioPacket) {
        self.tap.feed(&pkt.data);

        let mut buf = gst::Buffer::from_slice(pkt.data.clone());
        buf.get_mut()
            .unwrap()
            .set_pts(Some(gst::ClockTime::from_useconds(pkt.pts)));

        let _ = self.appsrc.push_buffer(buf);
    }
    pub fn finish(&mut self) {
        self.tap.flush();
        let _ = self.appsrc.end_of_stream();
    }
}