BREAKING CHANGE: Verbatim directives now sent to LLM for contextual splitting
- Keep (image: ...) and (video: ...) directives in prompt sent to LLM
- LLM instructed to include verbatim content in appropriate outputs
- Remove deterministic addition of verbatim content post-processing
- Add example in system prompt showing verbatim directive handling
- Better contextual understanding for LLM while maintaining precise control
- User request: don't remove verbatim text from LLM analysis
Benefits:
- LLM sees full context including verbatim requirements
- More intelligent splitting decisions based on complete prompt
- Maintains exact user control over content placement
- Simpler processing pipeline (less post-processing logic)
Oliver Switzer
fb01b596d6