The new Assistant API I mentioned refers to OpenAI’s Assistants v2 (or Threads/Run-based API). Instead of just chat completions, this lets you define assistants with tools, persistent memory, and instructions — and you interact through threads and runs. If you're using Vapi, they’ve been integrating with this newer version recently, so some older behaviors (like endOfCallMessages) don’t always work as expected. The new structure is more powerful but has some quirks when it comes to tools like ending a call. If you’re using Vapi’s built-in assistant config, make sure it’s compatible with the OpenAI Assistant v2 structure — and definitely keep an eye on their changelog because they’re evolving fast. Let me know if you need help debugging something specific — happy to check it out together!