Features

WhisperKit-CLI Integration: Native support for WhisperKit-CLI on Apple Silicon (5-10x faster transcription)
New reingest Command: Re-process specific episodes by ID to fix transcripts or test changes
- podsidian reingest <episode_id> [<episode_id> ...]
- Supports multiple episodes and --debug flag
- Full pipeline re-execution: download, transcribe, embed, export
Configurable Transcript Correction: New config options for transcript correction
- transcript_correction_enabled (default: false) - toggle LLM-based correction
- transcript_correction_chunk_size (default: 8000) - chunk size for long transcripts
Chunked Transcript Processing: Automatic chunking for long transcripts to prevent truncation
- Splits on sentence boundaries for natural breaks
- Processes each chunk separately and stitches results

Improvements

3-Tier Transcription Priority: RSS transcript → WhisperKit-CLI → Python Whisper
Automatic WhisperKit Detection: Seamlessly uses WhisperKit when available, falls back gracefully
Smart Model Path Resolution: Uses local WhisperKit models when available to avoid re-downloads
Enhanced Progress Feedback: Better status messages showing which transcription method is active
Cost Optimization: Transcript correction now disabled by default to reduce API costs

Fixed transcript truncation for long podcasts (2+ hours)
- Previously: ~5,000-10,000 characters due to LLM output limits
- Now: Full transcripts (130,000+ characters for 2-hour episodes)
Added max_tokens parameter to all OpenRouter API calls to prevent output truncation

Updated show-config command to display transcript correction settings
Added transcript correction options to config.toml.example with helpful documentation

Full Changelog: 1.0...1.1