When I look at the model output in my call logs it seem that llms messages arent truncated when the AI is interrupted. Wanted to confirm if this is the case ? Asking this since if messages arent truncated at any point in the conversation history then LLM will have zero knowledge that it got interrupted. Meaning it will have no idea if it has to repeat information the caller possibly didnt hear and will have a harder time handling interruptions more naturally.

Hey Alozie, the messages get truncated. when the next request is being sent

ok interesting it doesnt like ai has context of that as it responds like it already completed the previous sentence even though the user interrupted. Is there way to get an actual look at the conversational context vapi brings in for each response