feat: Added OpenAI.Microphone library and OpenAI.CLI to use Realtime API through console

OpenAI.sln

@@ -42,6 +42,10 @@ Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "workflows", "workflows", "{
 EndProject
 Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "GenerateDocs", "src\helpers\GenerateDocs\GenerateDocs.csproj", "{ECC219F0-209A-412B-ADEC-6D97AB379E7C}"
 EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "OpenAI.Microphone", "src\libs\OpenAI.Microphone\OpenAI.Microphone.csproj", "{C008F8CF-1A0E-4FE5-A5E1-5672D3B3A1E5}"
+EndProject
+Project("{FAE04EC0-301F-11D3-BF4B-00C04F79EFBC}") = "OpenAI.CLI", "src\libs\OpenAI.CLI\OpenAI.CLI.csproj", "{AAD812B2-77D6-42E0-A3FA-432BD6A73791}"
+EndProject
 Global
 	GlobalSection(SolutionConfigurationPlatforms) = preSolution
 		Debug|Any CPU = Debug|Any CPU
@@ -68,6 +72,14 @@ Global
 		{ECC219F0-209A-412B-ADEC-6D97AB379E7C}.Debug|Any CPU.Build.0 = Debug|Any CPU
 		{ECC219F0-209A-412B-ADEC-6D97AB379E7C}.Release|Any CPU.ActiveCfg = Release|Any CPU
 		{ECC219F0-209A-412B-ADEC-6D97AB379E7C}.Release|Any CPU.Build.0 = Release|Any CPU
+		{C008F8CF-1A0E-4FE5-A5E1-5672D3B3A1E5}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{C008F8CF-1A0E-4FE5-A5E1-5672D3B3A1E5}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{C008F8CF-1A0E-4FE5-A5E1-5672D3B3A1E5}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{C008F8CF-1A0E-4FE5-A5E1-5672D3B3A1E5}.Release|Any CPU.Build.0 = Release|Any CPU
+		{AAD812B2-77D6-42E0-A3FA-432BD6A73791}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{AAD812B2-77D6-42E0-A3FA-432BD6A73791}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{AAD812B2-77D6-42E0-A3FA-432BD6A73791}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{AAD812B2-77D6-42E0-A3FA-432BD6A73791}.Release|Any CPU.Build.0 = Release|Any CPU
 	EndGlobalSection
 	GlobalSection(SolutionProperties) = preSolution
 		HideSolutionNode = FALSE
@@ -79,6 +91,8 @@ Global
 		{A3F06E45-DFA8-4236-BFF5-425091762548} = {AAA11B78-2764-4520-A97E-46AA7089A588}
 		{2D8B78DE-7269-417B-9D0B-8981FA513ACB} = {E793AF18-4371-4EBD-96FC-195EB1798855}
 		{ECC219F0-209A-412B-ADEC-6D97AB379E7C} = {1A008ECD-2300-4BE4-A302-49DDF8BE0D54}
+		{C008F8CF-1A0E-4FE5-A5E1-5672D3B3A1E5} = {61E7E11E-4558-434C-ACE8-06316A3097B3}
+		{AAD812B2-77D6-42E0-A3FA-432BD6A73791} = {61E7E11E-4558-434C-ACE8-06316A3097B3}
 	EndGlobalSection
 	GlobalSection(ExtensibilityGlobals) = postSolution
 		SolutionGuid = {CED9A020-DBA5-4BE6-8096-75E528648EC1}

src/libs/OpenAI.CLI/Commands/VoiceCommand.cs

@@ -0,0 +1,28 @@
+using System.CommandLine;
+
+namespace OpenAI.CLI.Commands;
+
+/// <summary>
+/// 
+/// </summary>
+internal sealed class VoiceCommand : Command
+{
+    public VoiceCommand() : base(name: "voice", description: "Generates client sdk using a OpenAPI spec.")
+    {
+        var inputOption = new Argument<string>(
+            name: "input",
+            getDefaultValue: () => string.Empty,
+            description: "Input file path");
+        AddArgument(inputOption);
+
+        this.SetHandler(
+            HandleAsync,
+            inputOption);
+    }
+
+    private static async Task HandleAsync(
+        string input)
+    {
+        Console.WriteLine("Done.");
+    }
+}

src/libs/OpenAI.CLI/OpenAI.CLI.csproj

@@ -0,0 +1,25 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFramework>net8.0</TargetFramework>
+    <NoWarn>$(NoWarn)</NoWarn>
+  </PropertyGroup>
+
+  <PropertyGroup Label="NuGet">
+    <PackageId>tryAGI.OpenAI.CLI</PackageId>
+    <PackAsTool>true</PackAsTool>
+    <ToolCommandName>openai</ToolCommandName>
+    <Description>Advanced Voice from command line</Description>
+    <PackageTags>openai;advanced-voice;ai;realtime;api;microphone</PackageTags>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="System.CommandLine" Version="2.0.0-beta4.22272.1" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="..\OpenAI\OpenAI.csproj" />
+  </ItemGroup>
+
+</Project>

src/libs/OpenAI.CLI/Program.cs

@@ -0,0 +1,8 @@
+using System.CommandLine;
+using OpenAI.CLI.Commands;
+
+var rootCommand = new RootCommand(
+    description: "CLI tool to use AutoSDK");
+rootCommand.AddCommand(new VoiceCommand());
+
+return await rootCommand.InvokeAsync(args).ConfigureAwait(false);

src/libs/OpenAI.CLI/Properties/launchSettings.json

@@ -0,0 +1,8 @@
+{
+  "profiles": {
+    "FixOpenApiSpec": {
+      "commandName": "Project",
+      "commandLineArgs": ""
+    }
+  }
+}

src/libs/OpenAI.Microphone/Microphone.cs

@@ -0,0 +1,206 @@
+using System.Runtime.InteropServices;
+using PortAudioSharp;
+
+namespace OpenAI;
+
+/// <summary>
+/// Implements a Microphone using PortAudio
+/// </summary>
+public sealed class Microphone : IDisposable
+{
+    /// <summary>
+    /// 
+    /// </summary>
+    static Microphone()
+    {
+        PortAudio.Initialize();
+    }
+
+    /// <summary>
+    /// 
+    /// </summary>
+    public static void Terminate()
+    {
+        PortAudio.Terminate();
+    }
+
+    /// <summary>
+    /// 
+    /// </summary>
+    public Action<byte[], int>? PushCallback { get; set; }
+
+    /// <summary>
+    /// Sample rate
+    /// </summary>
+    public int SampleRate { get; set; } = 48_000;
+
+    /// <summary>
+    /// Chunk size
+    /// </summary>
+    public int ChunkSize { get; set; } = 32_000;
+
+    /// <summary>
+    /// Number of channels
+    /// </summary>
+    public int Channels { get; set; } = 1;
+
+    /// <summary>
+    /// Input device index
+    /// </summary>
+    public int DeviceIndex { get; set; } = PortAudio.NoDevice;
+
+    /// <summary>
+    /// Sample Format
+    /// </summary>
+    public SampleFormat SampleFormat { get; set; } = SampleFormat.Int16;
+
+    /// <summary>
+    /// 
+    /// </summary>
+    public bool IsMuted { get; private set; }
+
+    private PortAudioSharp.Stream? _stream;
+    private CancellationTokenSource? _exitToken;
+
+    /// <summary>
+    /// Start begins the listening on the microphone
+    /// </summary>
+    /// <returns></returns>
+    public bool Start()
+    {
+        if (_stream != null)
+        {
+            return false;
+        }
+
+        // reset exit token
+        _exitToken?.Dispose();
+        _exitToken = new CancellationTokenSource();
+
+        // Get the device info
+        if (DeviceIndex == PortAudio.NoDevice)
+        {
+            DeviceIndex = PortAudio.DefaultInputDevice;
+            if (DeviceIndex == PortAudio.NoDevice)
+            {
+                return false;
+            }
+        }
+
+        DeviceInfo info = PortAudio.GetDeviceInfo(DeviceIndex);
+
+        // Create the stream
+        _stream = new PortAudioSharp.Stream(
+            inParams: new StreamParameters
+            {
+                device = DeviceIndex,
+                channelCount = Channels,
+                sampleFormat = SampleFormat,
+                suggestedLatency = info.defaultLowInputLatency,
+                hostApiSpecificStreamInfo = IntPtr.Zero,
+            },
+            outParams: null,
+            sampleRate: SampleRate,
+            framesPerBuffer: (uint)ChunkSize,
+            streamFlags: StreamFlags.ClipOff,
+            callback: Сallback,
+            userData: IntPtr.Zero
+            );
+
+        // Start the stream
+        _stream.Start();
+        return true;
+    }
+
+    /// <summary>
+    /// 
+    /// </summary>
+    public void Stop()
+    {
+        // Check if we have a stream
+        if (_stream == null)
+        {
+            return;
+        }
+
+        // signal stop
+        _exitToken?.Cancel();
+
+        // Stop the stream
+        _stream.Stop();
+
+        _stream.Dispose();
+        _stream = null;
+        _exitToken?.Dispose();
+        _exitToken = null;
+    }
+
+    private StreamCallbackResult Сallback(
+        nint input,
+        nint output,
+        uint frameCount,
+        ref StreamCallbackTimeInfo timeInfo,
+        StreamCallbackFlags statusFlags,
+        nint userDataPtr)
+    {
+        // Check if the input is null
+        if (input == IntPtr.Zero)
+        {
+            return StreamCallbackResult.Continue;
+        }
+
+        // Check if the exit token is set
+        if (_exitToken is { IsCancellationRequested: true })
+        {
+            return StreamCallbackResult.Abort;
+        }
+
+        // copy and send the data
+        byte[] buf = new byte[frameCount * sizeof(Int16)];
+
+        if (IsMuted)
+        {
+            buf = new byte[buf.Length];
+        }
+        else
+        {
+            Marshal.Copy(input, buf, 0, buf.Length);
+        }
+
+        PushCallback?.Invoke(buf, buf.Length);
+
+        return StreamCallbackResult.Continue;
+    }
+
+    /// <summary>
+    /// 
+    /// </summary>
+    public void Mute()
+    {
+        if (_stream != null)
+        {
+            return;
+        }
+
+        IsMuted = true;
+    }
+
+    /// <summary>
+    /// 
+    /// </summary>
+    public void Unmute()
+    {
+        if (_stream != null)
+        {
+            return;
+        }
+
+        IsMuted = false;
+    }
+
+    /// <inheritdoc/>
+    public void Dispose()
+    {
+        Stop();
+    }
+}

src/libs/OpenAI.Microphone/OpenAI.Microphone.csproj

@@ -0,0 +1,24 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <TargetFrameworks>netstandard2.0;net8.0</TargetFrameworks>
+  </PropertyGroup>
+
+  <ItemGroup Label="CLSCompliant">
+    <AssemblyAttribute Include="System.CLSCompliantAttribute">
+      <_Parameter1>false</_Parameter1>
+    </AssemblyAttribute>
+  </ItemGroup>
+
+  <PropertyGroup Label="Nuget">
+    <PackageId>tryAGI.OpenAI.Microphone</PackageId>
+    <Description>Provides Microphone access</Description>
+    <PackageTags>openai;advanced-voice;ai;realtime;api;microphone</PackageTags>
+    <SignAssembly>false</SignAssembly>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="PortAudioSharp2" Version="1.0.0" />
+  </ItemGroup>
+
+</Project>

src/tests/OpenAI.IntegrationTests/OpenAI.IntegrationTests.csproj

@@ -19,6 +19,7 @@
     <PackageReference Include="FluentAssertions" Version="6.12.2" />
     <PackageReference Include="NUnit" Version="4.2.2" />
     <PackageReference Include="NUnit3TestAdapter" Version="4.6.0" />
+    <PackageReference Include="org.k2fsa.portaudio.runtime.osx-arm64" Version="1.0.4" />
   </ItemGroup>
 
   <ItemGroup Label="GlobalUsings">
@@ -30,6 +31,7 @@
   </ItemGroup>
 
   <ItemGroup>
+    <ProjectReference Include="..\..\libs\OpenAI.Microphone\OpenAI.Microphone.csproj" />
     <ProjectReference Include="..\..\libs\OpenAI\OpenAI.csproj" />
   </ItemGroup>
 

src/tests/OpenAI.IntegrationTests/Tests.Realtime.cs

@@ -2,6 +2,25 @@ namespace OpenAI.IntegrationTests;
 
 public partial class Tests
 {
+    [Test]
+    [Explicit]
+    public async Task Microphone()
+    {
+        using var microphone = new Microphone();
+        microphone.PushCallback = (bytes, count) =>
+        {
+            Console.WriteLine($"Pushed {count} bytes.");
+        };
+
+        microphone.Start();
+
+        await Task.Delay(TimeSpan.FromSeconds(15));
+
+        microphone.Stop();
+
+        OpenAI.Microphone.Terminate();
+    }
+
     [Test]
     [Explicit]
     public async Task Realtime()