During conversations, after the AI agent successfully calls the knowledge_bfq_tool query tool, it stops speaking entirely. No response or spoken output is generated from the KB result. The only way it resumes is if the user starts speaking again — then the agent continues as if the turn had never ended. Details: The tool call completes successfully and returns data. There is no TTS output or assistant turn triggered after the tool result. This was working perfectly about a week ago — no changes were made to models, prompts, or settings on our side. Suspect that the post-tool assistant turn is not being triggered automatically anymore. Question: Has anything changed recently in the way VAPI handles tool results → assistant response? What could cause the assistant to not resume speaking after a tool returns its data? Agent: https://agent.aicompass.com.au/na/prototype/XeymLzvDq5BFjpSE5f4y?_gl=1*19l2gv7*_gcl_au*MTc0OTY2ODY5NC4xNzU0NzM0MDI5

Recent updates require removing the request-complete message and ensuring synchronous tool responses in the correct format otherwise, the assistant may not speak tool results automatically after a call..