Connect Multi Agent Speech with the summaries created from book synth

[n3wang]

This should be working for having conversations,

It should also track the moment of the speech by filtering where not marked and marking while ongoing conversation with the person.