From 025b0a1d5cf3d187a71b6e5679dafb0c20d38e0c Mon Sep 17 00:00:00 2001
From: "nick.yi" <jqrddd@163.com>
Date: Wed, 12 Mar 2025 22:57:12 +0800
Subject: [PATCH] optimize input audio transcription

---
 .../Models/Realtime/RealtimeSessionBody.cs                  | 6 ++++++
 .../Providers/Realtime/RealTimeCompletionProvider.cs        | 5 +++--
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/src/Plugins/BotSharp.Plugin.OpenAI/Models/Realtime/RealtimeSessionBody.cs b/src/Plugins/BotSharp.Plugin.OpenAI/Models/Realtime/RealtimeSessionBody.cs
index 0d0947e5..ff2bfc3f 100644
--- a/src/Plugins/BotSharp.Plugin.OpenAI/Models/Realtime/RealtimeSessionBody.cs
+++ b/src/Plugins/BotSharp.Plugin.OpenAI/Models/Realtime/RealtimeSessionBody.cs
@@ -72,4 +72,10 @@ public class InputAudioTranscription
 {
     [JsonPropertyName("model")]
     public string Model { get; set; } = null!;
+
+    [JsonPropertyName("language")]
+    public string Language { get; set; } = "en";
+
+    [JsonPropertyName("prompt")]
+    public string Prompt { get; set; }
 }
\ No newline at end of file
diff --git a/src/Plugins/BotSharp.Plugin.OpenAI/Providers/Realtime/RealTimeCompletionProvider.cs b/src/Plugins/BotSharp.Plugin.OpenAI/Providers/Realtime/RealTimeCompletionProvider.cs
index 0d6f76b4..af6f0ad3 100644
--- a/src/Plugins/BotSharp.Plugin.OpenAI/Providers/Realtime/RealTimeCompletionProvider.cs
+++ b/src/Plugins/BotSharp.Plugin.OpenAI/Providers/Realtime/RealTimeCompletionProvider.cs
@@ -138,7 +138,7 @@ public class RealTimeCompletionProvider : IRealTimeCompletion
         {
             result = await _webSocket.ReceiveAsync(
                 new ArraySegment<byte>(buffer), CancellationToken.None);
-
+            
             // Convert received data to text/audio (Twilio sends Base64-encoded audio)
             string receivedText = Encoding.UTF8.GetString(buffer, 0, result.Count);
             if (string.IsNullOrEmpty(receivedText))
@@ -319,6 +319,7 @@ public class RealTimeCompletionProvider : IRealTimeCompletion
                 InputAudioTranscription = new InputAudioTranscription
                 {
                     Model = "whisper-1",
+                    Language = "en",
                 },
                 Voice = "alloy",
                 Instructions = instruction,
@@ -329,7 +330,7 @@ public class RealTimeCompletionProvider : IRealTimeCompletion
                 MaxResponseOutputTokens = 512,
                 TurnDetection = new RealtimeSessionTurnDetection
                 {
-                    Threshold = 0.8f,
+                    Threshold = 0.9f,
                     PrefixPadding = 300,
                     SilenceDuration = 800
                 }