Analyze this transcript recording and produce enrichment metadata as JSON.

## Recording Metadata
- Audio duration: {audio_duration_seconds} seconds ({audio_duration_minutes:.1f} minutes)
- Transcription confidence: {confidence:.1%}
- Language detected: {language_code}

## Chapters (from auto-transcription -- refine these)
{chapters_json}

## Transcript (speaker-attributed utterances)
{utterances_text}

## Speaker Identity
{speaker_identity_block}

## Names Roster (fallback context for speaker identification)
These are known people associated with the ashram. Use these only when the voiceprint identity above is unavailable for a label:
{names_roster}

## Controlled Topic Vocabulary (select ONLY from this list)
{topic_vocabulary}

Return the enrichment JSON now.
