feat: Add Groq STT provider integration

[Schreezer]

Summary

This PR adds comprehensive Groq STT (Speech-to-Text) provider integration to OpenSuperWhisper, bringing the fastest speech-to-text solution available to the app.

Key Features Added

• 🚀 Ultra-Fast Transcription: Groq provides 189-216x real-time speed transcription
• 🤖 Multiple Model Support:
  • whisper-large-v3-turbo: $0.04/hour, 12% WER, 216x speed (transcription only)
  • whisper-large-v3: $0.111/hour, 10.3% WER, 189x speed (transcription + translation)
• 🔒 Secure API Key Storage: Keychain integration for secure credential management
• ⚡ Advanced Features:
  • Word and segment-level timestamps
  • Multi-language support (30+ languages)
  • Custom prompts (up to 224 tokens)
  • Translation support (whisper-large-v3 only)
  • Comprehensive error handling and retry logic

Technical Implementation

Core Components

• GroqProvider.swift: Complete STT provider implementation with async/await
• Configuration: Integrated with existing STTTypes.swift and AppPreferences
• Security: API keys stored securely via SecureStorage system
• Factory Pattern: Integrated with STTProviderFactory for seamless provider switching

API Compliance

✅ Fully compliant with Groq STT API documentation

• OpenAI-compatible endpoints
• Proper multipart form-data requests
• Complete parameter support (model, language, prompt, response_format, temperature, timestamp_granularities)
• Comprehensive HTTP status code handling
• File size validation (25MB free tier, 100MB dev tier)

Quality & Performance

• Robust Error Handling: Comprehensive error mapping and user-friendly messages
• Retry Logic: Exponential backoff with configurable retry attempts
• File Validation: Audio format and size validation before upload
• Progress Tracking: Real-time progress callbacks for UI updates
• Memory Management: Efficient audio data handling for large files
• Network Optimization: Configurable timeouts and session management

Files Changed

• OpenSuperWhisper/STT/GroqProvider.swift (New): Complete Groq STT provider implementation
• OpenSuperWhisper/STT/STTTypes.swift: Added GroqConfiguration and provider type
• OpenSuperWhisper/STT/STTProviderFactory.swift: Integrated Groq provider creation
• OpenSuperWhisper/STT/SecureStorage.swift: Added secure Groq API key storage
• OpenSuperWhisper/Utils/AppPreferences.swift: Added Groq configuration preferences

Configuration Options

public struct GroqConfiguration {
    var endpoint: String = "https://api.groq.com/openai/v1/audio/transcriptions"
    var model: String = "whisper-large-v3-turbo" // Default to faster, cheaper model
    var maxRetries: Int = 3
    var timeoutInterval: TimeInterval = 60.0
    var maxFileSizeMB: Int = 25 // Free tier limit
    var apiKey: String? // Stored securely in Keychain
}

Usage

1. Setup: Add Groq API key in app settings
2. Provider Selection: Choose Groq as primary or fallback STT provider
3. Model Selection: Choose between turbo (faster/cheaper) or v3 (more accurate/translation)
4. Features: Enable timestamps, custom prompts, language specification

Testing

• API Integration: Comprehensive connectivity and authentication testing
• Error Scenarios: Complete error handling validation
• File Processing: Audio format and size validation
• Configuration: Secure storage and retrieval testing

Future Enhancements

• Translation Endpoint: Add dedicated translation endpoint support
• Audio Preprocessing: Client-side 16kHz mono conversion for optimal results
• Quality Metrics: Utilize response metadata (avg_logprob, compression_ratio) for quality monitoring
• Batch Processing: Support for large file chunking

Breaking Changes

None - this is a purely additive feature that maintains backward compatibility.

Performance Impact

• Positive: Groq provides the fastest STT processing available (189-216x real-time)
• Memory: Efficient streaming upload with minimal memory footprint
• Network: Optimized request/response handling with configurable timeouts

🤖 Generated with Claude Code

[madscientist111]

is there anyway for me to run this in my mac?

[Schreezer]

Refer this: https://github.com/Schreezer/STT