fix(podspec): sync version to v0.15.6 + CI guard against tag drift - #873
Conversation
The podspec self-reported 0.12.2 while tags advanced to v0.15.6, so any consumer resolving FluidAudio by version constraint gets a lie about what the tag contains (reported in react-native-fluidaudio#4 — the ~> 0.7 pin there can never reach the streaming managers, which first shipped in v0.15.3). Releases bump tags manually and nothing checked the podspec, so the string drifted silently for three minor versions. Bump spec.version to match the latest release and add a tag-push workflow that fails when the podspec at the tag disagrees with the tag name, so the mismatch surfaces at release time instead of in downstream Podfiles.
Supertonic3 Smoke Test ✅
Runtime: 0m25s Note: CI VMs lack a physical Neural Engine; the ANE-bucketed VectorEstimator falls back to CPU here. This validates download + variant resolution + synthesis, not ANE residency/perf. |
Parakeet EOU Benchmark Results ✅Status: Benchmark passed Performance Metrics
Streaming Metrics
Test runtime: 1m58s • 08/19/2026, 06:05 PM EST RTFx = Real-Time Factor (higher is better) • Processing includes: Model inference, audio preprocessing, state management, and file I/O |
VAD Benchmark ResultsPerformance Comparison
Dataset Details
✅: Average F1-Score above 70% |
Offline VBx Pipeline ResultsSpeaker Diarization Performance (VBx Batch Mode)Optimal clustering with Hungarian algorithm for maximum accuracy
Offline VBx Pipeline Timing BreakdownTime spent in each stage of batch diarization
Speaker Diarization Research ComparisonOffline VBx achieves competitive accuracy with batch processing
Pipeline Details:
🎯 Offline VBx Test • AMI Corpus ES2004a • 1049.0s meeting audio • 172.7s processing • Test runtime: 2m 58s • 08/19/2026, 06:12 PM EST |
PocketTTS Smoke Test ✅
Runtime: 0m26s Note: PocketTTS uses CoreML MLState (macOS 15) KV cache + Mimi streaming state. CI VM lacks physical GPU — audio quality and performance may differ from Apple Silicon. |
ASR Benchmark Results ✅Status: All benchmarks passed Parakeet v3 (multilingual)
Parakeet v2 (English-optimized)
Streaming (v3)
Streaming (v2)
Streaming tests use 5 files with 0.5s chunks to simulate real-time audio streaming 25 files per dataset • Test runtime: 10m25s • 08/19/2026, 06:15 PM EST RTFx = Real-Time Factor (higher is better) • Calculated as: Total audio duration ÷ Total processing time Expected RTFx Performance on Physical M1 Hardware:• M1 Mac: ~28x (clean), ~25x (other) Testing methodology follows HuggingFace Open ASR Leaderboard |
Speaker Diarization Benchmark ResultsSpeaker Diarization PerformanceEvaluating "who spoke when" detection accuracy
Diarization Pipeline Timing BreakdownTime spent in each stage of speaker diarization
Speaker Diarization Research ComparisonResearch baselines typically achieve 18-30% DER on standard datasets
Note: RTFx shown above is from GitHub Actions runner. On Apple Silicon with ANE:
🎯 Speaker Diarization Test • AMI Corpus ES2004a • 1049.0s meeting audio • 61.8s diarization time • Test runtime: 4m 25s • 08/19/2026, 06:15 PM EST |
Sortformer High-Latency Benchmark ResultsES2004a Performance (30.4s latency config)
Sortformer High-Latency • ES2004a • Runtime: 3m 29s • 2026-08-19T22:22:41.525Z |
1 participant
Summary
spec.versioninFluidAudio.podspecfrom 0.12.2 to 0.15.6 — the string hadn't been touched since the v0.12.2 release while tags advanced three minor versions, so consumers resolving by version constraint were told the tag contains something it doesn't (surfaced in feedStreamingAudio discards the audio it's given; FluidAudio pin predates the streaming managers react-native-fluidaudio#4).podspec-version.yml: on anyv*tag push, fail if the podspec version at that tag doesn't match the tag name, so the mismatch is caught at release time instead of in downstream Podfiles.Already-published tags keep the stale string (history is immutable); tag-only Podfile pins are unaffected either way. Starting with the next release the version string will be accurate again.
Test plan
pod ipc spec FluidAudio.podspecparses clean with the new version.0.15.6, matches tagv0.15.6.🤖 Generated with Claude Code