how can I use llama3 as openai to call

You can set your llm object with llama3 and define it globally with Settings. That way your defined llm will be used everywhere. Or you can pass llm as a kwarg too for generate_qa_embedding_pairs