add logging config to audio server

add audio client
change delimiter for mcp tools from "::" to ":" to save tokens

own_mcp/Cargo.toml

@@ -5,7 +5,7 @@ edition = "2024"
 
 [dependencies]
 ollama-rs = {version = "0.3.4", features = ["macros", "headers"]}
-reqwest = "0.13.2"
+reqwest = { version = "0.13.2", features = ["stream", "multipart", "form"] }
 tokio = { version = "1.50.0", features = ["rt", "rt-multi-thread", "macros"] }
 rmcp = {version="1.3.0", features = ["transport-streamable-http-client-reqwest", "reqwest", "client", "auth", "transport-child-process"]}
 log = {version = "0.4.29"}
@@ -13,4 +13,7 @@ env_logger = "0.11.10"
 serde = { version = "1.0.228", features = ["derive"] }
 thiserror = "2.0.17"
 url = "2.5.8"
-rand = "0.10.0"
+rand = "0.10.0"
+cpal = "0.17.3"
+hound = "3.5.1"
+bytes = "1.11.1"

own_mcp/src/audio/client.rs

@@ -0,0 +1,71 @@
+use std::io::ErrorKind;
+use std::path::Path;
+use reqwest::Client;
+use reqwest::multipart::Form;
+use thiserror::Error;
+use crate::audio::models::{AudioServerStatus, TranscriptionResponse, VoiceRequest};
+use url::Url;
+
+pub struct AudioClient{
+    client: Client,
+    authorization: Option<String>,
+    base_url: Url,
+}
+
+pub trait AudioClientTrait {
+    fn new(base_url: Url) -> Self;
+    fn from_client(base_url: Url, client: Client) -> Self;
+
+    fn with_authorization(self, authorization: String) -> Self;
+
+    fn status(&self) -> impl Future<Output = Result<AudioServerStatus, AudioError>>;
+
+    fn transcribe(&self, audio_file_path: impl AsRef<Path>) -> impl Future<Output=AudioResult<TranscriptionResponse>>;
+
+    fn tts(&self, voice_request: VoiceRequest) -> impl Future<Output=AudioResult<bytes::Bytes>>;
+}
+
+#[derive(Debug, Error)]
+pub enum AudioError {
+    #[error(transparent)]
+    RequestError(#[from] reqwest::Error),
+    #[error(transparent)]
+    UrlParseError(#[from] url::ParseError),
+    #[error(transparent)]
+    IOError(#[from] std::io::Error),
+}
+
+type AudioResult<T> = Result<T, AudioError>;
+
+impl AudioClientTrait for AudioClient {
+    fn new(base_url: Url) -> Self {
+        AudioClient { client: Client::new(), authorization: None, base_url }
+    }
+
+    fn from_client(base_url: Url, client: Client) -> Self {
+        AudioClient { client, authorization: None, base_url }
+    }
+
+    fn with_authorization(mut self, authorization: String) -> Self {
+        self.authorization = Some(authorization);
+        self
+    }
+
+    async fn status(&self) -> AudioResult<AudioServerStatus> {
+        let response = self.client.get(self.base_url.clone()).send().await?.error_for_status()?;
+        Ok(response.json().await?)
+    }
+
+    async fn transcribe(&self, audio_file_path: impl AsRef<Path>) -> AudioResult<TranscriptionResponse> {
+        let form = Form::new().file("audio_file", audio_file_path).await?;
+
+        let response = self.client.get(self.base_url.join("transcribe")?).multipart(form).send().await?;
+        Ok(response.json().await?)
+    }
+
+    async fn tts(&self, voice_request: VoiceRequest) -> AudioResult<bytes::Bytes> {
+        let response = self.client.get(self.base_url.join("tts")?).json(&voice_request).send().await?.error_for_status()?;
+
+        Ok(response.bytes().await?)
+    }
+}

own_mcp/src/audio/mod.rs

@@ -0,0 +1,4 @@
+pub mod recording;
+pub mod client;
+pub mod models;
+pub use client::{AudioClient, AudioClientTrait};

own_mcp/src/audio/models.rs

@@ -0,0 +1,60 @@
+use serde::{Deserialize, Serialize};
+
+#[derive(Debug, Clone, Deserialize)]
+pub struct AudioServerStatus {
+    #[serde(rename = "memory_usage")]
+    pub memory_usage_bytes: usize,
+    pub piper_model: String,
+    pub whisper_model: String,
+}
+
+//noinspection SpellCheckingInspection
+#[derive(Debug, Clone, Deserialize)]
+pub struct TranscriptionSegment {
+    pub id: u32,
+    pub seek: u32,
+    pub start: f32,
+    pub end: f32,
+    pub text: String,
+    pub tokens: Vec<u32>,
+    pub temperature: f32,
+    pub avg_logprob: f32,
+    pub compression_ratio: f32,
+    pub no_speech_prob: f32
+}
+
+#[derive(Debug, Clone, Deserialize)]
+pub struct TranscriptionResponse {
+    pub text: String,
+    pub segments: Vec<TranscriptionSegment>,
+    pub language: String
+}
+
+#[derive(Debug, Clone, Serialize)]
+pub struct SynthesisConfig {
+    pub speaker_id: Option<u32>,
+    pub length_scale: Option<u32>,
+    pub noise_scale: Option<u32>,
+    pub noise_w_scale: Option<u32>,
+    pub normalize_audio: bool,
+    pub volume: f32,
+}
+
+impl Default for SynthesisConfig {
+    fn default() -> Self {
+        SynthesisConfig {
+            speaker_id: None,
+            length_scale: None,
+            noise_scale: None,
+            noise_w_scale: None,
+            normalize_audio: true,
+            volume: 1.0,
+        }
+    }
+}
+
+#[derive(Debug, Clone, Serialize)]
+pub struct VoiceRequest {
+    pub text: String,
+    pub config: Option<SynthesisConfig>,
+}

own_mcp/src/audio/recording.rs

@@ -0,0 +1,197 @@
+use cpal::traits::{DeviceTrait, StreamTrait};
+use cpal::{BuildStreamError, Device, SampleFormat, SizedSample, Stream, SupportedStreamConfig};
+use std::fmt::Debug;
+use std::sync::{Arc, Mutex};
+use std::{path, thread};
+use std::thread::JoinHandle;
+use std::time::Duration;
+use hound::WavSpec;
+use log::{debug, error, info, trace};
+use thiserror::__private18::AsDisplay;
+use thiserror::Error;
+
+fn generic_sample_to_i16(sample: impl SizedSample + Into<f64>, format: SampleFormat) -> i16 {
+    match format {
+        SampleFormat::I8 => (sample.into() * 2.0) as i16,
+        SampleFormat::I16 => sample.into() as i16,
+        SampleFormat::I32 => (sample.into() / 2.0) as i16,
+        SampleFormat::F32 => (sample.into() * (i16::MAX as f64)) as i16,
+        _ => panic!("Unsupported sample format {:?}", format),
+    }
+}
+
+pub type SampleArc = Arc<Mutex<Vec<i16>>>;
+
+fn stream_callback<T: SizedSample + Debug + Into<f64>>(
+    input: &[T],
+    samples: SampleArc,
+    sample_format: SampleFormat,
+) {
+    if let Ok(mut guard) = samples.lock() {
+        for &sample in input {
+            guard.push(generic_sample_to_i16(sample, sample_format));
+        }
+    }
+}
+
+fn build_stream<T: SizedSample + Debug + hound::Sample + Send + Into<f64>>(
+    mic: &Device,
+    device_config: SupportedStreamConfig,
+    samples: SampleArc,
+    sample_format: SampleFormat,
+) -> Result<Stream, BuildStreamError> {
+    mic.build_input_stream(
+        &device_config.into(),
+        move |input: &[T], _info| stream_callback::<T>(input, samples.clone(), sample_format),
+        |e| {
+            error!("a stream error occurred while trying to record: {:?}", e);
+        },
+        None,
+    )
+}
+
+#[derive(Debug)]
+pub struct RecordingHandler {
+    pub samples: SampleArc,
+    pub spec: WavSpec,
+    thread_handle: JoinHandle<Result<(), RecordingError>>,
+    should_stop: Arc<Mutex<bool>>,
+}
+
+impl RecordingHandler {
+    pub fn stop_recording(self) -> Result<(), RecordingError> {
+        info!("Stopping recording");
+        *self.should_stop.lock().unwrap() = true;
+        self.thread_handle.join().unwrap()
+    }
+}
+
+#[derive(Error, Debug)]
+pub enum RecordingError {
+    #[error("unsupported sample format")]
+    UnsupportedSampleFormat(SampleFormat),
+    #[error(transparent)]
+    PlayStreamError(#[from] cpal::PlayStreamError),
+    #[error(transparent)]
+    BuildStreamError(#[from] BuildStreamError),
+    #[error("thread poisoned")]
+    ThreadPoison,
+    #[error("recording was already stopped")]
+    RecordingAlreadyStopped,
+}
+
+fn start_recording_blocking_with_parameters(
+    mic: &Device,
+    device_config: &SupportedStreamConfig,
+    sample_format: SampleFormat,
+    samples: SampleArc,
+    should_stop: Arc<Mutex<bool>>,
+) -> Result<(), RecordingError> {
+    let stream = match sample_format {
+        SampleFormat::I8 => {
+            build_stream::<i8>(mic, device_config.clone(), samples.clone(), sample_format)?
+        }
+        SampleFormat::I16 => {
+            build_stream::<i16>(mic, device_config.clone(), samples.clone(), sample_format)?
+        }
+        SampleFormat::I32 => {
+            build_stream::<i32>(mic, device_config.clone(), samples.clone(), sample_format)?
+        }
+        SampleFormat::F32 => {
+            build_stream::<f32>(mic, device_config.clone(), samples.clone(), sample_format)?
+        }
+        sample_format => {
+            return Err(RecordingError::UnsupportedSampleFormat(sample_format));
+        }
+    };
+
+    stream.play()?;
+
+    loop {
+        thread::sleep(Duration::from_millis(100));
+        if let Ok(guard) = should_stop.lock() {
+            if *guard {
+                debug!("Stopping recording gracefully");
+                break;
+            }
+        }
+    }
+
+    drop(stream);
+
+    Ok(())
+}
+
+pub fn start(
+    microphone: Device,
+    device_config: SupportedStreamConfig,
+) -> Arc<Mutex<Option<RecordingHandler>>> {
+    let sample_format = device_config.sample_format();
+
+    let samples: SampleArc = Arc::new(Mutex::new(Vec::new()));
+    let recording_should_stop = Arc::new(Mutex::new(false));
+
+    info!("started recording");
+    debug!("sample format: {:?}", sample_format);
+
+    let thread_samples = samples.clone();
+    let thread_recording_stop = recording_should_stop.clone();
+    let thread_device_config = device_config.clone();
+    let record_thread_handle = thread::spawn(move ||
+        start_recording_blocking_with_parameters(
+            &microphone,
+            &thread_device_config,
+            sample_format,
+            thread_samples,
+            thread_recording_stop,
+        )
+    );
+
+    let spec = WavSpec {
+        channels: device_config.channels(),
+        sample_rate: device_config.sample_rate(),
+        bits_per_sample: 16,
+        sample_format: hound::SampleFormat::Int,
+    };
+
+    debug!("device spec: {:?}", spec);
+
+    Arc::new(Mutex::new(Some(RecordingHandler {
+        spec,
+        samples,
+        thread_handle: record_thread_handle,
+        should_stop: recording_should_stop,
+    })))
+}
+
+pub fn stop_and_take_data(handler: Arc<Mutex<Option<RecordingHandler>>>) -> Result<(Vec<i16>, WavSpec), RecordingError> {
+    let handler = handler.lock().map_err(|_|RecordingError::ThreadPoison)?.take().ok_or(RecordingError::RecordingAlreadyStopped)?;
+
+    let samples = handler.samples.clone();
+    let spec = handler.spec.clone();
+
+    handler.stop_recording()?;
+
+    Ok((samples.lock().map_err(|_|RecordingError::ThreadPoison)?.to_vec(), spec))
+}
+
+pub fn samples_to_wav(
+    samples: impl IntoIterator<Item = i16>,
+    spec: &WavSpec,
+    filepath: impl AsRef<path::Path>
+) -> Result<Vec<u8>, hound::Error> {
+    let wav_bytes = std::io::Cursor::new(Vec::<u8>::new());
+
+    trace!("creating a wav file writer to {}", filepath.as_ref().display());
+    let mut file_writer = hound::WavWriter::create(&filepath, *spec)?;
+
+    trace!("writing samples...");
+    for sample in samples {
+        file_writer.write_sample(sample)?;
+    }
+
+    trace!("finalizing for file {}", filepath.as_ref().display());
+    file_writer.finalize()?;
+
+    Ok(wav_bytes.into_inner())
+}

own_mcp/src/lib.rs

@@ -1,3 +1,4 @@
 pub mod mcp;
+pub mod audio;
 
 pub use mcp::chat::AgentChat;

own_mcp/src/mcp/chat.rs

@@ -132,7 +132,7 @@ impl AgentChat {
     ///
     /// # Arguments
     ///
-    /// * `name`: must be in the format "mcp_server_name::tool_or_resource_name".
+    /// * `name`: must be in the format "mcp_server_name:tool_or_resource_name".
     ///
     /// returns:
     /// Ok(tuple) => tuple of the server name and the tool or resource name
@@ -141,12 +141,12 @@ impl AgentChat {
     /// # Examples
     ///
     /// ```ignore
-    /// let result = AgentChat::parse_tool_name("example::get_foo".to_string());
+    /// let result = AgentChat::parse_tool_name("example:get_foo".to_string());
     /// assert_eq!(result.unwrap(), ("example".to_string(), "get_foo".to_string()))
     /// ```
     fn parse_tool_name(name: String) -> Result<(String, String), ChatError> {
         let (mcp_server_name, tool_name) = name
-            .split_once("::")
+            .split_once(":")
             .ok_or(ChatError::FunctionParseError(name.clone()))?;
 
         Ok((mcp_server_name.to_string(), tool_name.to_string()))
@@ -161,7 +161,7 @@ impl AgentChat {
     /// # Arguments
     ///
     /// * `mpc_server_data`:
-    /// * `name`: must be in the format "mcp_server_name::tool_or_resource_name".
+    /// * `name`: must be in the format "mcp_server_name:tool_or_resource_name".
     ///
     /// returns: Option<&RestrictedTool>
     pub fn get_tool(
@@ -352,7 +352,7 @@ mod tests {
         assert_eq!(tools.len(), 1);
 
         assert_eq!(tools[0].permission, ToolPermission::Ask);
-        chat.set_permission("test::echo".to_string(), ToolPermission::Allowed).unwrap();
+        chat.set_permission("test:echo".to_string(), ToolPermission::Allowed).unwrap();
 
         let tools: Vec<RestrictedTool> = chat.get_all_tools().cloned().collect();
         assert_eq!(tools[0].permission, ToolPermission::Allowed);

own_mcp/src/mcp/translation.rs

@@ -32,7 +32,7 @@ fn tool_info_from_mcp_tool(mcp_tool: &rmcp::model::Tool, server_name: &String) -
     ToolInfo {
         tool_type: ToolType::Function,
         function: ToolFunctionInfo {
-            name: format!("{}::{}", server_name, mcp_tool.name),
+            name: format!("{}:{}", server_name, mcp_tool.name),
             description: mcp_tool
                 .description
                 .clone()
@@ -70,7 +70,7 @@ fn tool_info_from_mcp_resource(
     ToolInfo {
         tool_type: ToolType::Function,
         function: ToolFunctionInfo {
-            name: format!("{}::get_{}", server_name, mcp_resource.name),
+            name: format!("{}:get_{}", server_name, mcp_resource.name),
             description: format!(
                 "type: {} - {}",
                 mcp_resource