v0.7.4

Don't make BGN playbooks public by default

Update message batching config and update packages

Avoid recursion in playbooks attribute check

fix: LangFuse spans for message do not show message

fix: Say() calls inside generated if and else blocks are streamed

fix: LangFuse LLM calls do not have token counts

Refactor token counting to use tiktoken and utility functions

Refine trigger evaluation and yield examples in prompts

Revise final instructions in InterpreterPrompt

Implement unified message batching for agents

Return value is optional

Update README.md

Update readme

Stop tracking .llm_cache_test folder