LE prompt with reasoning

app.py

@@ -551,11 +551,7 @@ Intelligent Thinking Context: {thinking_json}
 Your task is to generate the social media clips following these strict rules:
 
 1. TIMESTAMP SELECTION:
-- You must check the transcript line by line
-- Verify speaker continuity with NO interruptions
-- End clips immediately before any other speaker starts
-- If Speaker A talks from 1:00-1:10, then Speaker B talks, then Speaker A resumes at 1:15, these must be separate clips
-- Never combine timestamps across interruptions
+- Make sure that a selected timestamp range of a clip has a single speaker only. We need to isolate each speaker's CLIPs.
 
 2. CLIP REQUIREMENTS:
 - Minimum 20 seconds of CONTINUOUS speech