Update tokio-websockets in the azure-speech-sdk-rs beause of a potential websocket upgrade problem and add a timeout of 5 seconds to all Azure connects

Cargo.toml

@@ -22,7 +22,7 @@ members = [
 ]
 
 [workspace.package]
-version = "3.4.0"
+version = "3.4.1"
 edition = "2024"
 license = "MIT"
 repository = "https://github.com/pragmatrix/context-switch"

services/azure/src/lib.rs

@@ -1,3 +1,8 @@
+use std::{future::Future, time::Duration};
+
+use anyhow::{Context, Result};
+use tokio::time::timeout;
+
 mod host;
 // TODO: Attempt to make the modules non-pub
 pub mod synthesize;
@@ -9,3 +14,13 @@ pub use host::*;
 pub use synthesize::AzureSynthesize;
 pub use transcribe::AzureTranscribe;
 pub use translate::AzureTranslate;
+
+// The 256-frame input buffer fills in about 5.1 seconds at the usual 20 ms frame cadence.
+// Fail the connection first so buffer overflow does not terminate the conversation instead.
+const CONNECT_TIMEOUT: Duration = Duration::from_secs(5);
+
+async fn connect_with_timeout<T>(connection: impl Future<Output = T>) -> Result<T> {
+    timeout(CONNECT_TIMEOUT, connection)
+        .await
+        .context("Azure connection timed out")
+}

services/azure/src/synthesize.rs

@@ -14,7 +14,7 @@ use context_switch_core::{
     AudioFrame, BillingRecord, BillingSchedule, Conversation, Input, Service,
 };
 
-use crate::Host;
+use crate::{Host, connect_with_timeout};
 
 #[derive(Debug, Serialize, Deserialize)]
 #[serde(rename_all = "camelCase")]
@@ -64,7 +64,8 @@ impl Service for AzureSynthesize {
             .enable_session_end()
             .with_audio_format(azure_audio_format);
 
-        let client = synthesizer::Client::connect(host.auth.clone(), config).await?;
+        let client =
+            connect_with_timeout(synthesizer::Client::connect(host.auth.clone(), config)).await??;
 
         let language = params.language;
         let (mut input, output) = conversation.start()?;

services/azure/src/transcribe.rs

@@ -13,7 +13,7 @@ use context_switch_core::{
     speech_gate::make_speech_gate_processor_soft_rms,
 };
 
-use crate::Host;
+use crate::{Host, connect_with_timeout};
 
 #[derive(Debug, Deserialize)]
 #[serde(rename_all = "camelCase")]
@@ -90,7 +90,8 @@ impl Service for AzureTranscribe {
         }
         .set_output_format(recognizer::OutputFormat::Detailed);
 
-        let client = recognizer::Client::connect(host.auth.clone(), config).await?;
+        let client =
+            connect_with_timeout(recognizer::Client::connect(host.auth.clone(), config)).await??;
 
         let (mut input, output) = conversation.start()?;
 

services/azure/src/translate.rs

@@ -6,7 +6,7 @@ use futures::StreamExt;
 use serde::{Deserialize, Serialize};
 use tracing::{debug, error};
 
-use crate::Host;
+use crate::{Host, connect_with_timeout};
 use context_switch_core::{
     AudioFormat, AudioFrame, BillingRecord, BillingSchedule, Conversation, Input, OutputModality,
     OutputPath, Service,
@@ -74,7 +74,8 @@ impl Service for AzureTranslate {
             }
         };
 
-        let client = translator::Client::connect(host.auth.clone(), config).await?;
+        let client =
+            connect_with_timeout(translator::Client::connect(host.auth.clone(), config)).await??;
 
         let (mut input, output) = conversation.start()?;