v1.0.0

Dictation works end to end (spec 0001, on
spec 0002).

Added

• Streaming transcription through a whisper-flow server (1.1.0+): /ready check, version
  check, WebSocket streaming of 16 kHz PCM, and finalizing the last phrase on stop (1.2.0+).
• Microphone capture in a sidecar process: SoX, arecord or ffmpeg, or
  whisperCode.recorderCommand.
• Live editor insertion: partial text is dimmed and replaced as you speak, each phrase is one
  undo step, and the selection is replaced.
• Workspace vocabulary: identifiers from open editors are sent as Whisper's prompt (1.2.0+)
  and respelled in the output.
• Context-aware formatting for code comments and commit messages.
• Settings whisperCode.serverUrl, whisperCode.model, whisperCode.vocabulary.enabled and
  whisperCode.recorderCommand; commands Set API Key and Clear API Key (secret
  storage).
• Confirmation before sending audio to a non-local server, and a warning without TLS.
• Status bar states: idle, connecting, recording, error. Errors for a bad API key, a full
  server, and lost connections (with Retry).
• Comment wrapping (whisperCode.wrap.comments, default 80) with the comment marker carried to
  each line, and commit body wrapping (whisperCode.wrap.commitBody, default 72).
• Vocabulary from workspace source files that aren't open (whisperCode.vocabulary.scanWorkspace,
  whisperCode.vocabulary.maxFiles).
• Dictation into the AI chat input: the shortcut, pressed in the chat input, types each final
  phrase there. VS Code doesn't expose the chat input to extensions as a text editor.
• A manual E2E workflow that runs dictation against whisper-flow in Docker with a recorded
  WAV (npm run e2e locally).
• README header with the Whisper Code logo.

Changed

• The 95% coverage gate measures extension code only; tests and fakes no longer count.

Security

• whisperCode.recorderCommand can only be set in user (machine) settings, not in a
  workspace's .vscode/settings.json, so opening a repository can't make dictation run a
  command it chose.

v0.1.0

First preview release. Sets up the extension and project tooling; audio capture and
transcription are not implemented yet.

Added

• Whisper Code: Toggle Dictation command, bound to Cmd+Alt+D
  (macOS) / Ctrl+Alt+D (Windows/Linux).
• Status bar item that shows whether dictation is on and toggles it when clicked.
• Design spec 0001: Overview.
• CI: formatting, type-aware linting, type checking, tests with a 95% coverage gate,
  VSIX packaging, and CodeQL analysis.