You’re on the right track with splitting by language, but instead of duplicating assistants inside one squad, I’d structure it as language-based routing at the entry point. Set up a lightweight “router” assistant (or webhook layer via Make.com) that detects language first, then directs the call to a dedicated squad (English squad vs Vietnamese squad). That way, each squad stays clean and maintainable without duplication. For Vietnamese specifically, I’d also recommend testing a dedicated STT like Google Speech-to-Text or Whisper instead of multilingual mode—they’re typically more consistent. If you want, I can help you design the routing flow so it scales without turning into a mess. @Pedro Lourenco