v0.4.0

Release v0.4.0

• Add an interactive Bubble Tea TUI for one-off local transcription jobs.
• Add API key management commands with macOS Keychain support and local key-file fallback.
• Use one API key resolution order across CLI, pipeline, doctor, and TUI flows.
• Refactor CLI spec building into a shared tested package used by CLI and TUI.
• Improve TUI option editing, selection controls, output-path hints, and API key configuration screens.
• Expand gitignore coverage for generated transcripts, local configs, scratch data, and private agent notes.

v0.3.0

Release v0.3.0

• Add --parallel client-chunk sending for gpt-4o-transcribe and gpt-4o-mini-transcribe.
• Keep sequential execution as the default and emit explicit warnings when parallel mode disables prompt carryover.
• Record client chunk execution mode in manifests and require --resume runs to match the original sequential vs. parallel mode.
• Treat --parallel as a no-op for single-request or server-chunked plans rather than failing the job.
• Add resumable client-chunked transcription via manifest-backed chunk caches.
• Default --retries to 1 for retryable provider failures and reset input streams correctly before retry.
• Refresh README and contract/config documentation for parallel chunking and resume behavior.

v0.2.0

Release v0.2.0

• Switch ffmpeg-generated intermediate audio from large WAV files to compact provider-ready .m4a artifacts.
• Keep workdir artifacts aligned with what was actually uploaded to the provider.
• Add a single-request runtime guard that rejects normalized audio exceeding the provider upload size limit.
• Update intermediate-size estimation to match compressed AAC/M4A output more closely.

v0.1.1

Release v0.1.1

• Fix GitHub Actions lint execution so CI works correctly with the repository Go version.
• Package Windows release artifacts as .zip archives with a .exe binary inside.
• Keep macOS and Linux release packaging unchanged as .tar.gz.