If you're using a single multimodal model like 4o vapi will be a place for squads, tools, integrations, and deployments if you're using not a multimodal model it's a place to do that and string together the llm tts and transcription technologies with low latency