Realtime inference API for open-source models in your agent: two LLMs, text-to-speech and speech-to-text. Try the latest open-source models on your own prompts, add a key to your code, speak text and transcribe audio.