Move Rule Engine LLM to a smaller LLM

[ErezSha]

In order to reduce latency, try moving the rule engine LLM to a smaller & faster LLM (like gpt-4o)

1. see how it works.
2. check the latency is cut down

[ErezSha]

gpt-3.5 should be enough for the entry point node gpt-4o should be enough for rule engine node