Enable semantic voice activity detection option (#545)

ShivangiReja · web-flow · commit aa0baf8cc737 · 2025-07-21T11:43:22.000-07:00
Enable semantic voice activity detection option
diff --git a/api/OpenAI.net8.0.cs b/api/OpenAI.net8.0.cs
@@ -3828,6 +3828,23 @@ public class ResponseStartedUpdate : RealtimeUpdate, IJsonModel<ResponseStartedU
         protected override BinaryData PersistableModelWriteCore(ModelReaderWriterOptions options);
     }
     [Experimental("OPENAI002")]
+    public readonly partial struct SemanticEagernessLevel : IEquatable<SemanticEagernessLevel> {
+        public SemanticEagernessLevel(string value);
+        public static SemanticEagernessLevel Auto { get; }
+        public static SemanticEagernessLevel High { get; }
+        public static SemanticEagernessLevel Low { get; }
+        public static SemanticEagernessLevel Medium { get; }
+        public readonly bool Equals(SemanticEagernessLevel other);
+        [EditorBrowsable(EditorBrowsableState.Never)]
+        public override readonly bool Equals(object obj);
+        [EditorBrowsable(EditorBrowsableState.Never)]
+        public override readonly int GetHashCode();
+        public static bool operator ==(SemanticEagernessLevel left, SemanticEagernessLevel right);
+        public static implicit operator SemanticEagernessLevel(string value);
+        public static bool operator !=(SemanticEagernessLevel left, SemanticEagernessLevel right);
+        public override readonly string ToString();
+    }
+    [Experimental("OPENAI002")]
     public class TranscriptionSessionConfiguredUpdate : RealtimeUpdate, IJsonModel<TranscriptionSessionConfiguredUpdate>, IPersistableModel<TranscriptionSessionConfiguredUpdate> {
         public RealtimeContentModalities ContentModalities { get; }
         public RealtimeAudioFormat InputAudioFormat { get; }
@@ -3862,6 +3879,7 @@ public enum TurnDetectionKind {
     public class TurnDetectionOptions : IJsonModel<TurnDetectionOptions>, IPersistableModel<TurnDetectionOptions> {
         public TurnDetectionKind Kind { get; }
         public static TurnDetectionOptions CreateDisabledTurnDetectionOptions();
+        public static TurnDetectionOptions CreateSemanticVoiceActivityTurnDetectionOptions(SemanticEagernessLevel? eagernessLevel = null, bool? enableAutomaticResponseCreation = null, bool? enableResponseInterruption = null);
         public static TurnDetectionOptions CreateServerVoiceActivityTurnDetectionOptions(float? detectionThreshold = null, TimeSpan? prefixPaddingDuration = null, TimeSpan? silenceDuration = null, bool? enableAutomaticResponseCreation = null, bool? enableResponseInterruption = null);
         protected virtual TurnDetectionOptions JsonModelCreateCore(ref Utf8JsonReader reader, ModelReaderWriterOptions options);
         protected virtual void JsonModelWriteCore(Utf8JsonWriter writer, ModelReaderWriterOptions options);
diff --git a/api/OpenAI.netstandard2.0.cs b/api/OpenAI.netstandard2.0.cs
@@ -3370,6 +3370,22 @@ public class ResponseStartedUpdate : RealtimeUpdate, IJsonModel<ResponseStartedU
         protected override RealtimeUpdate PersistableModelCreateCore(BinaryData data, ModelReaderWriterOptions options);
         protected override BinaryData PersistableModelWriteCore(ModelReaderWriterOptions options);
     }
+    public readonly partial struct SemanticEagernessLevel : IEquatable<SemanticEagernessLevel> {
+        public SemanticEagernessLevel(string value);
+        public static SemanticEagernessLevel Auto { get; }
+        public static SemanticEagernessLevel High { get; }
+        public static SemanticEagernessLevel Low { get; }
+        public static SemanticEagernessLevel Medium { get; }
+        public readonly bool Equals(SemanticEagernessLevel other);
+        [EditorBrowsable(EditorBrowsableState.Never)]
+        public override readonly bool Equals(object obj);
+        [EditorBrowsable(EditorBrowsableState.Never)]
+        public override readonly int GetHashCode();
+        public static bool operator ==(SemanticEagernessLevel left, SemanticEagernessLevel right);
+        public static implicit operator SemanticEagernessLevel(string value);
+        public static bool operator !=(SemanticEagernessLevel left, SemanticEagernessLevel right);
+        public override readonly string ToString();
+    }
     public class TranscriptionSessionConfiguredUpdate : RealtimeUpdate, IJsonModel<TranscriptionSessionConfiguredUpdate>, IPersistableModel<TranscriptionSessionConfiguredUpdate> {
         public RealtimeContentModalities ContentModalities { get; }
         public RealtimeAudioFormat InputAudioFormat { get; }
@@ -3401,6 +3417,7 @@ public enum TurnDetectionKind {
     public class TurnDetectionOptions : IJsonModel<TurnDetectionOptions>, IPersistableModel<TurnDetectionOptions> {
         public TurnDetectionKind Kind { get; }
         public static TurnDetectionOptions CreateDisabledTurnDetectionOptions();
+        public static TurnDetectionOptions CreateSemanticVoiceActivityTurnDetectionOptions(SemanticEagernessLevel? eagernessLevel = null, bool? enableAutomaticResponseCreation = null, bool? enableResponseInterruption = null);
         public static TurnDetectionOptions CreateServerVoiceActivityTurnDetectionOptions(float? detectionThreshold = null, TimeSpan? prefixPaddingDuration = null, TimeSpan? silenceDuration = null, bool? enableAutomaticResponseCreation = null, bool? enableResponseInterruption = null);
         protected virtual TurnDetectionOptions JsonModelCreateCore(ref Utf8JsonReader reader, ModelReaderWriterOptions options);
         protected virtual void JsonModelWriteCore(Utf8JsonWriter writer, ModelReaderWriterOptions options);
diff --git a/scripts/Export-Api.ps1 b/scripts/Export-Api.ps1
@@ -258,6 +258,9 @@ function Invoke-GenAPI {
     # Remove Diagnostics.DebuggerStepThrough attribute.
     $content = $content -creplace ".*Diagnostics.DebuggerStepThrough.*\n", ""
 
+    # Remove ModelReaderWriterBuildable attributes.
+    $content = $content -creplace '\[ModelReaderWriterBuildable\(typeof\([^\)]+\)\)\]\s*', ''
+
     # Remove internal APIs.
     $content = $content -creplace "  * internal.*`n", ""
 
diff --git a/src/Custom/Realtime/GeneratorStubs.cs b/src/Custom/Realtime/GeneratorStubs.cs
@@ -11,3 +11,4 @@ namespace OpenAI.Realtime;
 [CodeGenType("RealtimeResponseUsage")] public partial class ConversationTokenUsage { }
 [CodeGenType("RealtimeToolType")] public readonly partial struct ConversationToolKind { }
 [CodeGenType("DotNetRealtimeVoiceIds")] public readonly partial struct ConversationVoice { }
+[CodeGenType("RealtimeSemanticVadTurnDetectionEagerness")] public readonly partial struct SemanticEagernessLevel { }
diff --git a/src/Custom/Realtime/Internal/GeneratorStubs.cs b/src/Custom/Realtime/Internal/GeneratorStubs.cs
@@ -35,7 +35,6 @@ namespace OpenAI.Realtime;
 [Experimental("OPENAI002")][CodeGenType("RealtimeResponseStatusDetailsType")] internal readonly partial struct InternalRealtimeResponseStatusDetailsType { }
 [Experimental("OPENAI002")][CodeGenType("RealtimeResponseVoice")] internal readonly partial struct InternalRealtimeResponseVoice { }
 [Experimental("OPENAI002")][CodeGenType("RealtimeSemanticVadTurnDetection")] internal partial class InternalRealtimeSemanticVadTurnDetection { }
-[Experimental("OPENAI002")][CodeGenType("RealtimeSemanticVadTurnDetectionEagerness")] internal readonly partial struct InternalRealtimeSemanticVadTurnDetectionEagerness { }
 [Experimental("OPENAI002")][CodeGenType("RealtimeServerEventConversationCreated")] internal partial class InternalRealtimeServerEventConversationCreated { }
 [Experimental("OPENAI002")][CodeGenType("RealtimeServerEventConversationCreatedConversation")] internal partial class InternalRealtimeServerEventConversationCreatedConversation { }
 [Experimental("OPENAI002")][CodeGenType("RealtimeServerEventConversationItemInputAudioTranscriptionFailedError")] internal partial class InternalRealtimeServerEventConversationItemInputAudioTranscriptionFailedError { }
diff --git a/src/Custom/Realtime/TurnDetectionOptions.Serialization.cs b/src/Custom/Realtime/TurnDetectionOptions.Serialization.cs
@@ -20,6 +20,7 @@ internal static TurnDetectionOptions DeserializeTurnDetectionOptions(JsonElement
             {
                 case "none": return InternalRealtimeNoTurnDetection.DeserializeInternalRealtimeNoTurnDetection(element, options);
                 case "server_vad": return InternalRealtimeServerVadTurnDetection.DeserializeInternalRealtimeServerVadTurnDetection(element, options);
+                case "semantic_vad": return InternalRealtimeSemanticVadTurnDetection.DeserializeInternalRealtimeSemanticVadTurnDetection(element, options);
                 default: return null;
             }
         }
diff --git a/src/Custom/Realtime/TurnDetectionOptions.cs b/src/Custom/Realtime/TurnDetectionOptions.cs
@@ -35,8 +35,8 @@ public static TurnDetectionOptions CreateServerVoiceActivityTurnDetectionOptions
         };
     }
 
-    internal static TurnDetectionOptions CreateSemanticVoiceActivityTurnDetectionOptions(
-        InternalRealtimeSemanticVadTurnDetectionEagerness? eagernessLevel = null,
+    public static TurnDetectionOptions CreateSemanticVoiceActivityTurnDetectionOptions(
+        SemanticEagernessLevel? eagernessLevel = null,
         bool? enableAutomaticResponseCreation = null,
         bool? enableResponseInterruption = null)
     {
diff --git a/src/Generated/Models/Realtime/InternalRealtimeSemanticVadTurnDetection.Serialization.cs b/src/Generated/Models/Realtime/InternalRealtimeSemanticVadTurnDetection.Serialization.cs
@@ -57,7 +57,7 @@ internal static InternalRealtimeSemanticVadTurnDetection DeserializeInternalReal
             bool? responseCreationEnabled = default;
             bool? responseInterruptionEnabled = default;
             IDictionary<string, BinaryData> additionalBinaryDataProperties = new ChangeTrackingDictionary<string, BinaryData>();
-            InternalRealtimeSemanticVadTurnDetectionEagerness? eagerness = default;
+            SemanticEagernessLevel? eagerness = default;
             foreach (var prop in element.EnumerateObject())
             {
                 if (prop.NameEquals("type"u8))
@@ -89,7 +89,7 @@ internal static InternalRealtimeSemanticVadTurnDetection DeserializeInternalReal
                     {
                         continue;
                     }
-                    eagerness = new InternalRealtimeSemanticVadTurnDetectionEagerness(prop.Value.GetString());
+                    eagerness = new SemanticEagernessLevel(prop.Value.GetString());
                     continue;
                 }
                 // Plugin customization: remove options.Format != "W" check
diff --git a/src/Generated/Models/Realtime/InternalRealtimeSemanticVadTurnDetection.cs b/src/Generated/Models/Realtime/InternalRealtimeSemanticVadTurnDetection.cs
@@ -13,11 +13,11 @@ public InternalRealtimeSemanticVadTurnDetection() : this(TurnDetectionKind.Seman
         {
         }
 
-        internal InternalRealtimeSemanticVadTurnDetection(TurnDetectionKind kind, bool? responseCreationEnabled, bool? responseInterruptionEnabled, IDictionary<string, BinaryData> additionalBinaryDataProperties, InternalRealtimeSemanticVadTurnDetectionEagerness? eagerness) : base(kind, responseCreationEnabled, responseInterruptionEnabled, additionalBinaryDataProperties)
+        internal InternalRealtimeSemanticVadTurnDetection(TurnDetectionKind kind, bool? responseCreationEnabled, bool? responseInterruptionEnabled, IDictionary<string, BinaryData> additionalBinaryDataProperties, SemanticEagernessLevel? eagerness) : base(kind, responseCreationEnabled, responseInterruptionEnabled, additionalBinaryDataProperties)
         {
             Eagerness = eagerness;
         }
 
-        internal InternalRealtimeSemanticVadTurnDetectionEagerness? Eagerness { get; set; }
+        public SemanticEagernessLevel? Eagerness { get; set; }
     }
 }
diff --git a/src/Generated/Models/Realtime/InternalRealtimeSemanticVadTurnDetectionEagerness.cs b/src/Generated/Models/Realtime/InternalRealtimeSemanticVadTurnDetectionEagerness.cs
diff --git a/src/Generated/Models/Realtime/SemanticEagernessLevel.cs b/src/Generated/Models/Realtime/SemanticEagernessLevel.cs
@@ -0,0 +1,52 @@
+// <auto-generated/>
+
+#nullable disable
+
+using System;
+using System.ComponentModel;
+using System.Diagnostics.CodeAnalysis;
+using OpenAI;
+
+namespace OpenAI.Realtime
+{
+    [Experimental("OPENAI002")]
+    public readonly partial struct SemanticEagernessLevel : IEquatable<SemanticEagernessLevel>
+    {
+        private readonly string _value;
+        private const string LowValue = "low";
+        private const string MediumValue = "medium";
+        private const string HighValue = "high";
+        private const string AutoValue = "auto";
+
+        public SemanticEagernessLevel(string value)
+        {
+            Argument.AssertNotNull(value, nameof(value));
+
+            _value = value;
+        }
+
+        public static SemanticEagernessLevel Low { get; } = new SemanticEagernessLevel(LowValue);
+
+        public static SemanticEagernessLevel Medium { get; } = new SemanticEagernessLevel(MediumValue);
+
+        public static SemanticEagernessLevel High { get; } = new SemanticEagernessLevel(HighValue);
+
+        public static SemanticEagernessLevel Auto { get; } = new SemanticEagernessLevel(AutoValue);
+
+        public static bool operator ==(SemanticEagernessLevel left, SemanticEagernessLevel right) => left.Equals(right);
+
+        public static bool operator !=(SemanticEagernessLevel left, SemanticEagernessLevel right) => !left.Equals(right);
+
+        public static implicit operator SemanticEagernessLevel(string value) => new SemanticEagernessLevel(value);
+
+        [EditorBrowsable(EditorBrowsableState.Never)]
+        public override bool Equals(object obj) => obj is SemanticEagernessLevel other && Equals(other);
+
+        public bool Equals(SemanticEagernessLevel other) => string.Equals(_value, other._value, StringComparison.InvariantCultureIgnoreCase);
+
+        [EditorBrowsable(EditorBrowsableState.Never)]
+        public override int GetHashCode() => _value != null ? StringComparer.InvariantCultureIgnoreCase.GetHashCode(_value) : 0;
+
+        public override string ToString() => _value;
+    }
+}
diff --git a/tests/Realtime/RealtimeSmokeTests.cs b/tests/Realtime/RealtimeSmokeTests.cs
@@ -198,6 +198,18 @@ public void TurnDetectionSerializationWorks()
         JsonNode serializedNode = JsonNode.Parse(serializedOptions);
         Assert.That(serializedNode["turn_detection"]?["type"]?.GetValue<string>(), Is.EqualTo("server_vad"));
         Assert.That(serializedNode["turn_detection"]?["threshold"]?.GetValue<float>(), Is.EqualTo(0.42f));
+
+        sessionOptions = new()
+        {
+            TurnDetectionOptions = TurnDetectionOptions.CreateSemanticVoiceActivityTurnDetectionOptions(
+                SemanticEagernessLevel.Medium, true, true)
+        };
+        serializedOptions = ModelReaderWriter.Write(sessionOptions);
+        serializedNode = JsonNode.Parse(serializedOptions);
+        Assert.That(serializedNode["turn_detection"]?["type"]?.GetValue<string>(), Is.EqualTo("semantic_vad"));
+        Assert.That(serializedNode["turn_detection"]?["eagerness"]?.GetValue<string>(), Is.EqualTo("medium"));
+        Assert.That(serializedNode["turn_detection"]?["create_response"]?.GetValue<bool>(), Is.EqualTo(true));
+        Assert.That(serializedNode["turn_detection"]?["interrupt_response"]?.GetValue<bool>(), Is.EqualTo(true));
     }
 
     [Test]

Original file line number	Diff line number	Diff line change
`@@ -20,6 +20,7 @@ internal static TurnDetectionOptions DeserializeTurnDetectionOptions(JsonElement`
`20`	`20`	`{`
`21`	`21`	`case "none": return InternalRealtimeNoTurnDetection.DeserializeInternalRealtimeNoTurnDetection(element, options);`
`22`	`22`	`case "server_vad": return InternalRealtimeServerVadTurnDetection.DeserializeInternalRealtimeServerVadTurnDetection(element, options);`
	`23`	`+ case "semantic_vad": return InternalRealtimeSemanticVadTurnDetection.DeserializeInternalRealtimeSemanticVadTurnDetection(element, options);`
`23`	`24`	`default: return null;`
`24`	`25`	`}`
`25`	`26`	`}`
Original file line number	Diff line number	Diff line change
`@@ -35,8 +35,8 @@ public static TurnDetectionOptions CreateServerVoiceActivityTurnDetectionOptions`
`35`	`35`	`};`
`36`	`36`	`}`
`37`	`37`
`38`		`- internal static TurnDetectionOptions CreateSemanticVoiceActivityTurnDetectionOptions(`
`39`		`- InternalRealtimeSemanticVadTurnDetectionEagerness? eagernessLevel = null,`
	`38`	`+ public static TurnDetectionOptions CreateSemanticVoiceActivityTurnDetectionOptions(`
	`39`	`+ SemanticEagernessLevel? eagernessLevel = null,`
`40`	`40`	`bool? enableAutomaticResponseCreation = null,`
`41`	`41`	`bool? enableResponseInterruption = null)`
`42`	`42`	`{`
Original file line number	Diff line number	Diff line change
`@@ -57,7 +57,7 @@ internal static InternalRealtimeSemanticVadTurnDetection DeserializeInternalReal`
`57`	`57`	`bool? responseCreationEnabled = default;`
`58`	`58`	`bool? responseInterruptionEnabled = default;`
`59`	`59`	`IDictionary<string, BinaryData> additionalBinaryDataProperties = new ChangeTrackingDictionary<string, BinaryData>();`
`60`		`- InternalRealtimeSemanticVadTurnDetectionEagerness? eagerness = default;`
	`60`	`+ SemanticEagernessLevel? eagerness = default;`
`61`	`61`	`foreach (var prop in element.EnumerateObject())`
`62`	`62`	`{`
`63`	`63`	`if (prop.NameEquals("type"u8))`
`@@ -89,7 +89,7 @@ internal static InternalRealtimeSemanticVadTurnDetection DeserializeInternalReal`
`89`	`89`	`{`
`90`	`90`	`continue;`
`91`	`91`	`}`
`92`		`- eagerness = new InternalRealtimeSemanticVadTurnDetectionEagerness(prop.Value.GetString());`
	`92`	`+ eagerness = new SemanticEagernessLevel(prop.Value.GetString());`
`93`	`93`	`continue;`
`94`	`94`	`}`
`95`	`95`	`// Plugin customization: remove options.Format != "W" check`
Original file line number	Diff line number	Diff line change
`@@ -13,11 +13,11 @@ public InternalRealtimeSemanticVadTurnDetection() : this(TurnDetectionKind.Seman`
`13`	`13`	`{`
`14`	`14`	`}`
`15`	`15`
`16`		`- internal InternalRealtimeSemanticVadTurnDetection(TurnDetectionKind kind, bool? responseCreationEnabled, bool? responseInterruptionEnabled, IDictionary<string, BinaryData> additionalBinaryDataProperties, InternalRealtimeSemanticVadTurnDetectionEagerness? eagerness) : base(kind, responseCreationEnabled, responseInterruptionEnabled, additionalBinaryDataProperties)`
	`16`	`+ internal InternalRealtimeSemanticVadTurnDetection(TurnDetectionKind kind, bool? responseCreationEnabled, bool? responseInterruptionEnabled, IDictionary<string, BinaryData> additionalBinaryDataProperties, SemanticEagernessLevel? eagerness) : base(kind, responseCreationEnabled, responseInterruptionEnabled, additionalBinaryDataProperties)`
`17`	`17`	`{`
`18`	`18`	`Eagerness = eagerness;`
`19`	`19`	`}`
`20`	`20`
`21`		`- internal InternalRealtimeSemanticVadTurnDetectionEagerness? Eagerness { get; set; }`
	`21`	`+ public SemanticEagernessLevel? Eagerness { get; set; }`
`22`	`22`	`}`
`23`	`23`	`}`