How argmaxinc/argmax-oss-swift works
Argmax Open-Source SDK packages three on-device inference workflows: WhisperKit transcription, SpeakerKit diarization, and TTSKit speech generation, with a CLI and local transcription server as user-facing entry points. Shared ArgmaxCore handles model resolution/loading and tokenization. The graph follows documented workflows and relationships named by the SOURCE INDEX; unsampled implementation details are not expanded into speculative call chains.
Main entry points: WhisperKit.swift WhisperKit · SpeakerKit.swift SpeakerKit · TTSKit.swift TTSKit · TranscribeCLI.swift Transcribe command
Components 21
User interfaces
- Argmax CLI
Sources/ArgmaxCLI/ArgmaxCLI.swift - Transcribe command
Sources/ArgmaxCLI/TranscribeCLI.swift - Local API server
Sources/ArgmaxCLI/Server/GeneratedSources/Server.swift - OpenAI handler
Sources/ArgmaxCLI/Server/OpenAIHandler.swift - TTS command
Sources/ArgmaxCLI/TTSCLI.swift
Shared model services
- Model manager
Sources/ArgmaxCore/ModelManager.swift - Model downloader
Sources/ArgmaxCore/ModelDownloader.swift - Hub API
Sources/ArgmaxCore/External/Hub/HubApi.swift - Hub configuration
Sources/ArgmaxCore/External/Hub/Config.swift - Tokenizer
Sources/ArgmaxCore/External/Tokenizers/Tokenizer.swift
Transcription
- WhisperKit
Sources/WhisperKit/Core/WhisperKit.swift - Audio processing
Sources/WhisperKit/Core/Audio/AudioProcessor.swift - Text decoding
Sources/WhisperKit/Core/TextDecoder.swift - Transcription results
Sources/WhisperKit/Utilities/TranscriptionUtilities.swift
Speaker diarization
- SpeakerKit
Sources/SpeakerKit/SpeakerKit.swift - Pyannote models
Sources/SpeakerKit/Pyannote/PyannoteModelManager.swift - Diarization results
Sources/SpeakerKit/DiarizationResult.swift
Speech generation
- TTSKit
Sources/TTSKit/TTSKit.swift - Qwen generation
Sources/TTSKit/Qwen3TTS/Qwen3GenerateTask.swift - Audio output
Sources/TTSKit/Utilities/AudioOutput.swift
Other components
- App or user
Connections 26
- App or user → Argmax CLI: runs
README.md - Argmax CLI → Transcribe command: dispatches
Sources/ArgmaxCLI/ArgmaxCLI.swift - App or user → Local API server: sends requests
README.md - Local API server → OpenAI handler: routes to
- OpenAI handler → WhisperKit: transcribes with
Sources/ArgmaxCLI/Server/OpenAIHandler.swift - Transcribe command → WhisperKit: transcribes with
Sources/ArgmaxCLI/TranscribeCLI.swift - Transcribe command → SpeakerKit: diarizes with
Sources/ArgmaxCLI/TranscribeCLI.swift - WhisperKit → Audio processing: loads audio
Sources/WhisperKit/Core/WhisperKit.swift - WhisperKit → Text decoding: decodes text
Sources/WhisperKit/Core/WhisperKit.swift - WhisperKit → Transcription results: produces results
Sources/WhisperKit/Core/WhisperKit.swift - SpeakerKit → Pyannote models: loads models
Sources/SpeakerKit/Pyannote/PyannoteModelManager.swift - SpeakerKit → Diarization results: returns diarization
Sources/SpeakerKit/SpeakerKit.swift - SpeakerKit → Transcription results: aligns speakers
- App or user → TTS command: runs
README.md - TTS command → TTSKit: generates with
Sources/ArgmaxCLI/TTSCLI.swift - TTS command → Audio output: plays or saves
Sources/ArgmaxCLI/TTSCLI.swift - TTSKit → Qwen generation: generates speech
Sources/TTSKit/TTSKit.swift - TTSKit → Audio output: plays audio
Sources/TTSKit/TTSKit.swift - Model manager → Model downloader: resolves models
Sources/ArgmaxCore/ModelManager.swift - Model downloader → Hub API: fetches model files
Sources/ArgmaxCore/ModelDownloader.swift - Hub API → Hub configuration: uses configuration
Sources/ArgmaxCore/External/Hub/HubApi.swift - WhisperKit → Model manager: manages model state
- SpeakerKit → Model manager: uses model manager
Sources/SpeakerKit/SpeakerKit.swift - TTSKit → Model downloader: downloads models
Sources/TTSKit/TTSKit.swift - WhisperKit → Tokenizer: loads tokenizer
Sources/WhisperKit/Core/WhisperKit.swift - TTSKit → Tokenizer: loads tokenizer
Sources/TTSKit/TTSKit.swift