Merge pull request #85 from pragmatrix/vl-noise-suppression

Interface Voice Live NoiseReductionType, so that everything is serialized in cam…

Author: pragmatrix

services/microsoft-voice-live/src/client.rs

@@ -20,7 +20,7 @@ use context_switch_core::{
     Input, OutputPath, ThresholdLevel, audio,
 };
 
-use crate::transcribe::{Params, ServiceOutputEvent};
+use crate::transcribe::{NoiseReduction, NoiseReductionType, Params, ServiceOutputEvent};
 use crate::transcription_state::TranscriptionState;
 
 pub struct Client {
@@ -114,7 +114,7 @@ impl Client {
     async fn send_session_update(&mut self, params: &Params) -> Result<()> {
         let session = types::VoiceLiveSession {
             input_audio_sampling_rate: None,
-            input_audio_noise_reduction: params.noise_reduction.clone(),
+            input_audio_noise_reduction: params.noise_reduction.as_ref().map(noise_reduction),
             input_audio_echo_cancellation: None,
             input_audio_transcription: Some(types::TranscriptionConfig {
                 language: params.language.clone(),
@@ -321,6 +321,18 @@ fn eou_threshold_level(level: ThresholdLevel) -> EndOfUtteranceThresholdLevel {
     }
 }
 
+fn noise_reduction(configured: &NoiseReduction) -> types::NoiseReduction {
+    let reduction_type = match configured.reduction_type {
+        NoiseReductionType::NearField => types::NoiseReductionType::NearField,
+        NoiseReductionType::FarField => types::NoiseReductionType::FarField,
+        NoiseReductionType::AzureDeepNoiseSuppression => {
+            types::NoiseReductionType::AzureDeepNoiseSuppression
+        }
+    };
+
+    types::NoiseReduction { reduction_type }
+}
+
 enum FlowControl {
     Continue,
     End,

services/microsoft-voice-live/src/transcribe.rs

@@ -1,6 +1,5 @@
 use anyhow::Result;
 use async_trait::async_trait;
-use openai_api_rs::realtime::types::NoiseReduction;
 use serde::{Deserialize, Serialize};
 
 use context_switch_core::{Conversation, Service, TurnDetection};
@@ -36,6 +35,26 @@ pub struct Params {
     pub turn_detection: Option<TurnDetection>,
 }
 
+/// Input-audio noise reduction. Mapped to the provider noise-reduction configuration before
+/// being sent in the session update.
+#[derive(Debug, Deserialize)]
+pub struct NoiseReduction {
+    #[serde(rename = "type")]
+    pub reduction_type: NoiseReductionType,
+}
+
+/// Noise-reduction profile. Mirrors the provider's variants with a camelCase wire form.
+#[derive(Debug, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub enum NoiseReductionType {
+    /// `nearField` is for close-talking microphones such as headphones.
+    NearField,
+    /// `farField` is for far-field microphones such as laptop or conference room microphones.
+    FarField,
+    /// Azure deep noise suppression, optimized for the speaker closest to the microphone.
+    AzureDeepNoiseSuppression,
+}
+
 #[derive(Debug)]
 pub struct MicrosoftVoiceLiveTranscribe;