BotSharp/src/Plugins/BotSharp.Plugin.AudioHandler/Provider/NativeWhisperProvider.cs

115 lines
3.5 KiB
C#
Raw Normal View History

2024-07-30 18:57:32 +00:00
using Whisper.net;
using Whisper.net.Ggml;
namespace BotSharp.Plugin.AudioHandler.Provider;
/// <summary>
/// Native Whisper provider for speech to text conversion
/// </summary>
2024-08-28 16:08:12 +00:00
public class NativeWhisperProvider : IAudioCompletion
2024-07-30 18:57:32 +00:00
{
2024-08-26 22:24:07 +00:00
private static WhisperProcessor _whisperProcessor;
2024-08-28 16:08:12 +00:00
private readonly IServiceProvider _services;
private readonly IFileStorageService _fileStorage;
2024-08-26 22:24:07 +00:00
private readonly ILogger<NativeWhisperProvider> _logger;
public string Provider => "native";
2024-07-30 18:57:32 +00:00
2024-08-11 16:26:57 +00:00
public NativeWhisperProvider(
2024-08-28 16:08:12 +00:00
BotSharpDatabaseSettings dbSettings,
IFileStorageService fileStorage,
IServiceProvider services,
2024-08-11 16:26:57 +00:00
ILogger<NativeWhisperProvider> logger)
2024-07-30 18:57:32 +00:00
{
2024-08-28 16:08:12 +00:00
_fileStorage = fileStorage;
_services = services;
2024-08-11 16:26:57 +00:00
_logger = logger;
2024-07-30 18:57:32 +00:00
}
2024-08-28 16:08:12 +00:00
public async Task<string> GenerateTextFromAudioAsync(Stream audio, string audioFileName, string? text = null)
2024-07-30 18:57:32 +00:00
{
var textResult = new List<SegmentData>();
2024-08-28 16:08:12 +00:00
using var stream = AudioHelper.Transform(audio, audioFileName);
2024-08-26 22:24:07 +00:00
await foreach (var result in _whisperProcessor.ProcessAsync(stream).ConfigureAwait(false))
2024-07-30 18:57:32 +00:00
{
textResult.Add(result);
}
2024-08-26 22:24:07 +00:00
_whisperProcessor.Dispose();
2024-08-28 16:08:12 +00:00
stream.Close();
2024-08-17 23:34:31 +00:00
2024-07-30 18:57:32 +00:00
var audioOutput = new AudioOutput
{
Segments = textResult
};
return audioOutput.ToString();
}
2024-08-26 22:24:07 +00:00
2024-08-28 18:43:37 +00:00
public async Task<BinaryData> GenerateAudioFromTextAsync(string text)
2024-08-26 22:24:07 +00:00
{
throw new NotImplementedException();
}
2024-08-28 16:08:12 +00:00
public void SetModelName(string model)
2024-08-26 22:24:07 +00:00
{
if (Enum.TryParse(model, true, out GgmlType ggmlType))
{
2024-08-28 16:08:12 +00:00
LoadWhisperModel(ggmlType);
}
else
{
_logger.LogWarning($"Unsupported model type: {model}. Use Tiny model instead!");
LoadWhisperModel(GgmlType.Tiny);
2024-08-26 22:24:07 +00:00
}
}
2024-08-28 16:08:12 +00:00
private void LoadWhisperModel(GgmlType modelType)
2024-07-30 18:57:32 +00:00
{
try
{
2024-08-28 16:08:12 +00:00
var modelDir = _fileStorage.BuildDirectory("models", "whisper");
var exist = _fileStorage.ExistDirectory(modelDir);
if (!exist)
2024-08-26 22:24:07 +00:00
{
2024-08-28 16:08:12 +00:00
_fileStorage.CreateDirectory(modelDir);
2024-08-26 22:24:07 +00:00
}
2024-07-30 18:57:32 +00:00
2024-08-28 16:08:12 +00:00
var files = _fileStorage.GetFiles("models/whisper", "*.bin");
var modelLoc = files.FirstOrDefault(x => Path.GetFileName(x) == BuildModelFile(modelType));
if (string.IsNullOrEmpty(modelLoc))
2024-08-17 23:34:31 +00:00
{
2024-08-28 16:08:12 +00:00
modelLoc = BuildModelPath(modelType);
DownloadModel(modelType, modelLoc);
2024-07-30 18:57:32 +00:00
}
2024-08-17 23:34:31 +00:00
2024-08-28 16:08:12 +00:00
var bytes = _fileStorage.GetFileBytes(modelLoc);
_whisperProcessor = WhisperFactory.FromBuffer(buffer: bytes).CreateBuilder().WithLanguage("auto").Build();
2024-07-30 18:57:32 +00:00
}
catch (Exception ex)
{
2024-08-26 22:24:07 +00:00
var error = "Failed to load whisper model";
_logger.LogWarning($"${error}: {ex.Message}\r\n{ex.InnerException}");
throw new Exception($"{error}: {ex.Message}");
2024-07-30 18:57:32 +00:00
}
}
2024-08-28 16:08:12 +00:00
private void DownloadModel(GgmlType modelType, string modelDir)
{
using var modelStream = WhisperGgmlDownloader.GetGgmlModelAsync(modelType).ConfigureAwait(false).GetAwaiter().GetResult();
_fileStorage.SaveFileStreamToPath(modelDir, modelStream);
modelStream.Close();
}
private string BuildModelPath(GgmlType modelType)
2024-07-30 18:57:32 +00:00
{
2024-08-28 16:08:12 +00:00
return _fileStorage.BuildDirectory("models", "whisper", BuildModelFile(modelType));
2024-07-30 18:57:32 +00:00
}
2024-08-11 16:26:57 +00:00
2024-08-28 16:08:12 +00:00
private string BuildModelFile(GgmlType modelType)
2024-08-17 23:34:31 +00:00
{
2024-08-28 16:08:12 +00:00
return $"ggml-{modelType}.bin";
2024-08-17 23:34:31 +00:00
}
2024-07-30 18:57:32 +00:00
}