In the MeetingCopilot UI, how does the AudioLines binding work to segment audio into speaker lines, and how is the active speaker highlighted in the transcript log?

In the MeetingCopilot UI, the AudioLines binding is the bridge between the raw audio stream and the transcript view. When the Whisper model emits a chunk of text, AudioLines parses the timestamped tokens and groups them by speaker using the diarization output. Each group becomes a speaker line object that the UI renders in the transcript log. The orange bg-orange-500/5 highlight is applied to the line whose speaker ID matches the current speaker context, which is updated in real time as the audio progresses. This visual cue lets users quickly identify who is speaking. The binding also exposes a selection API so that clicking a line sets the active speaker state, which in turn updates the highlight and can trigger actions like fetching speaker‑specific metadata from the knowledge base. By keeping the binding logic separate from the UI components, we can swap out Whisper or change the diarization algorithm without touching the rendering code.