tomi/hush
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

add Fireworks AI provider support

Browse Source
This commit is contained in:
Tomi Eckert
2026-03-22 19:02:37 +01:00
parent a26d453720
commit 70e784a1cc
4 changed files with 310 additions and 73 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Hush.Daemon/src/Orchestrator.cs
+29 -35
View File
@@ -8,11 +8,10 @@ namespace Hush.Daemon;
public class Orchestrator
{
private static readonly HttpClient _httpClient = new();
private readonly ConfigManager _configManager;
private readonly IAudioRecorder _recorder;
private IAudioToTextProvider? _audioToTextProvider;
private ITextStreamingProvider? _textProvider;
private ITextInput? _textInput;
private string? _recordingPath;
private DateTime? _recordingStartTime;
@@ -149,9 +148,19 @@ public class Orchestrator
var provider = GetTextProvider(config);
var prompt = $"""
Process this spoken text for clarity and correctness. Fix any errors, add proper punctuation, and make it read naturally. Keep the original meaning intact.
You are a transcription post-processor. Your task is to clean up raw speech-to-text output and return polished, ready-to-type text.
Text: {text}
Rules:
- Detect the language of the transcription and process it entirely in that language — do not translate
- Fix grammar, spelling, and punctuation errors introduced by the speech recognizer, following the conventions of the detected language
- Capitalize sentences and proper nouns appropriately for the detected language
- Remove filler words and false starts appropriate to the detected language (e.g. "um", "uh", "like" in English; "euh", "bah" in French; "äh", "ähm" in German; "eh", "tipo" in Spanish/Italian)
- Preserve the speaker's original intent, vocabulary choices, and tone
- Do not add, remove, or reinterpret content beyond what was said
- Do not include any explanation, preamble, or metadata — output only the corrected text
- If the input is empty or unintelligible, return an empty string
Raw transcription: {text}
""";
return await provider.CompleteTextAsync(prompt, config.LlmModel);
@@ -163,51 +172,36 @@ public class Orchestrator
await input.TypeString(text);
}
private IAudioToTextProvider GetAudioToTextProvider(HushConfig config)
{
if (_audioToTextProvider != null)
return _audioToTextProvider;
_audioToTextProvider = config.WhisperProvider switch
private IAudioToTextProvider GetAudioToTextProvider(HushConfig config) =>
config.WhisperProvider switch
{
"groq" => string.IsNullOrEmpty(config.GroqApiKey)
"groq" => string.IsNullOrEmpty(config.GroqApiKey)
? throw new InvalidOperationException("Groq API key is required for Whisper transcription")
: new GroqProvider(config.GroqApiKey),
: new GroqProvider(config.GroqApiKey, _httpClient),
"fireworks" => string.IsNullOrEmpty(config.FireworksApiKey)
? throw new InvalidOperationException("Fireworks API key is required for Whisper transcription")
: new FireworksProvider(config.FireworksApiKey, _httpClient),
_ => throw new InvalidOperationException($"Unsupported Whisper provider: {config.WhisperProvider}")
};
return _audioToTextProvider;
}
private ITextStreamingProvider GetTextProvider(HushConfig config)
{
if (_textProvider != null)
return _textProvider;
_textProvider = config.LlmProvider switch
private ITextStreamingProvider GetTextProvider(HushConfig config) =>
config.LlmProvider switch
{
"groq" => string.IsNullOrEmpty(config.GroqApiKey)
? throw new InvalidOperationException("Groq API key is required for LLM")
: new GroqProvider(config.GroqApiKey),
: new GroqProvider(config.GroqApiKey, _httpClient),
"fireworks" => string.IsNullOrEmpty(config.FireworksApiKey)
? throw new InvalidOperationException("Fireworks API key is required for LLM")
: new FireworksProvider(config.FireworksApiKey, _httpClient),
_ => throw new InvalidOperationException($"Unsupported LLM provider: {config.LlmProvider}")
};
return _textProvider;
}
private ITextInput GetTextInput(HushConfig config)
{
if (_textInput != null)
return _textInput;
_textInput = config.TypingBackend switch
private static ITextInput GetTextInput(HushConfig config) =>
config.TypingBackend switch
{
"xdotool" => new XdotoolInput(),
_ => new WtypeInput()
};
return _textInput;
}
private IAudioRecorder CreateAudioRecorder()
{