Hey chat, today I have around 50 assistants distributed across my clients. We have generated 5 different tools and all of them use those 5. "send_text", "send_image", "send_audio". When my client start configuring the tool, in my frontend he put those tools information (Image file, audio mp4...) and my database saves it with a unique identificator When we add that tool inside our assistant, we ask it to call the tool (send_text for exemple), and we order it to put the tool_id inside the arguments. But suddently our assistant just stopped putting the id inside the arguments, which broke all my tools. Does somebody suffered from the same problem? Im wondering how to fix that, or change our structure so all my clients creates their own tools (I can create tools snapshot, but my fear is any limitations on how many tools my vapi account can have) Any help?

Yeah, I’ve seen this before, sometimes the model just stops strictly following the tool schema, especially if tool_id isn’t clearly marked as required. If that ID is essential, I’d enforce it at the schema level and also add backend validation so any tool call without it just fails safely. Relying only on the model to always include it can be risky. Creating separate tools per client could work, but it might add a lot of overhead long term. Did anything change recently, model version, temperature, or the way the tool schema/prompt is structured? @zucchi0657

Hi, could you please provide us with some of the affected call IDs?