OSSSpeechKit 1.0: Modern async API, iOS 17, SPM-only distribution

[AppDevGuy]

OSSSpeechKit 1.0: Modern async API, iOS 17, SPM-only distribution

Pull Request Notes

Summary

This PR ships OSSSpeechKit 1.0.0, a major modernization of the library. It introduces an instance-based, async/await API centered on OSSSpeechEngine, replaces subclass-based voice/utterance configuration with value types, adds a runtime-aware language catalog, and moves distribution exclusively to Swift Package Manager with an iOS 17 minimum deployment target.

Legacy APIs (OSSSpeech.shared, OSSVoice, OSSUtterance, delegate-based recognition) remain available but are deprecated to allow incremental migration.

What's New

• OSSSpeechEngine — Main-actor-isolated engine for speech synthesis and recognition with typed OSSSpeechError failures
• Async recognition — recognitionEvents(locale:) returns an AsyncThrowingStream of OSSSpeechRecognitionEvent (partial, completed, availability, cancelled)
• Continuous dictation — continuousRecognitionEvents(locale:) with pauseRecognition(), resumeRecognition(), and finishRecognition() for long-running sessions with segment timestamps
• OSSLanguage catalog — Display metadata, runtime synthesis voice lookup, recognition support checks, and Unicode flagEmoji / UIKit renderedFlagImage(pointSize:)
• Value-type configuration — OSSVoiceConfiguration and OSSUtteranceConfiguration replace OSSVoice / OSSUtterance subclasses
• Privacy manifest — PrivacyInfo.xcprivacy bundled with the package
• Example app — New recording session UI (RecordingSessionViewController, RecordingSessionModel) demonstrating continuous recognition
• Docs & migration — Updated README.md, new MIGRATION.md and CHANGELOG.md, regenerated Jazzy API docs

Breaking Changes

Area	Before (0.x)	After (1.0)
Minimum iOS	Earlier targets	iOS 17+
Installation	CocoaPods + SPM	SPM only (historical 0.3.3 pod remains resolvable but unsupported)
API style	OSSSpeech.shared singleton + delegates	Instance-based OSSSpeechEngine with async/await
Voice selection	Silent fallback to unavailable voices	Throws OSSSpeechError.voiceUnavailable
Recognition	Delegate callbacks (didFinishListening, etc.)	Async event streams; call cancelRecognition() to stop
Language/flags	OSSVoiceEnum.allCases + image flags	OSSLanguage.catalog + flagEmoji / rendered images

Deprecated (still present)

• OSSSpeech / OSSSpeech.shared
• OSSVoice / OSSUtterance
• OSSSpeechDelegate and delegate-based recognition
• OSSVoiceEnum catalog and image flag access

See MIGRATION.md for side-by-side migration examples.

Infrastructure

• Swift tools version bumped to 6.0; package targets iOS 17
• CocoaPods podspec and Example Podfile removed
• CI moved to updated GitHub Actions workflow; Travis config removed
• Example project links the local package directly (no CocoaPods workspace)

Test Plan

• swift test passes (package tests under Example/Tests)
• Example app builds and runs on iOS 17+ device/simulator
• Speech synthesis works for a catalog language with installed voices
• One-shot recognition (recognitionEvents) produces partial and final transcripts
• Continuous recognition session: start → pause → resume → finish
• Permission denial surfaces as OSSSpeechError (not silent failure)
• Voice unavailable for locale throws OSSSpeechError.voiceUnavailable
• Legacy OSSSpeech API still compiles with deprecation warnings
• SPM integration from a clean Xcode project resolves 1.0.0