my agent is triggering a "role": "tool_calls", but he answer the question before received the tool_call_result (ex call ID : 7de5313b-9a93-44cb-a777-2ca7833206cc) so almost all the time he answer with the wrong answer. how to fix that ? i've already tried to add - When you call "tool_calls", wait "tool_call_result" before answering in the prompt. but it doesnt fix but it seems something that VAPI need to support nativly

Hi there! 👋 This is a known issue with VAPI agents. Let me share some configuration options that might help: 1. **Use synchronous tool calls**: Make sure your tool is configured with

async: false

to ensure the function waits for the completion of the tool execution before proceeding [Discord conversation](https://discord.com/channels/1211482211119796234/1211483291191083018/1333880245681393756). 2. **Configure blocking messages**: Try using

ToolMessageStart.blocking=true

which is a recent update that allows you to configure your tool calls to wait until a message is fully spoken before starting [Changelog](https://docs.vapi.ai/changelog/2025/2/4). 3. **Check response format**: Ensure your server responds with the correct JSON format containing

toolCallId

and `result`:

json
{
  "results": [
    {
      "toolCallId": "X",
      "result": "Y"
    }
  ]
}

There are some known limitations: - Some users have reported that VAPI doesn't always wait for tool_call_result before proceeding - With async settings, the client might return a default 'Success' message - For Realtime API, requests might proceed with a placeholder message that a tool result is still pending You're right that this might need better native support from VAPI, as several users have reported similar issues in the Discord community. Let me know if you need more specific help with your implementation! Sources: - [Discord conversation about tool call issues](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1342336047589359618) - [VAPI Tools Calling documentation](https://docs.vapi.ai/tools-calling) - [VAPI Changelog](https://docs.vapi.ai/changelog/2025/2/4) - [Discord conversation about async/sync tool calls](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1333880245681393756)

where should i put ToolMessageStart.blocking = true ??

I don't have the specific implementation details for where to place

ToolMessageStart.blocking = true

in your code based on the information I found. 🤔 This feature appears to be a recent update mentioned in the [VAPI Changelog](https://docs.vapi.ai/changelog/2025/2/4), but the exact implementation isn't detailed in the information I have access to. I'd recommend: 1. Checking the full changelog documentation for implementation examples 2. Looking at the VAPI tools documentation for more detailed guidance 3. If you're using an SDK, check if there's a specific method or property for setting this configuration Would you like me to try finding more specific information about this implementation detail? Source: - [VAPI Changelog](https://docs.vapi.ai/changelog/2025/2/4)

@Solaris could you please provide more specific details in the following format?  To help track down this issue, could you share: - What response you expected to get - What response you actually got instead This would really help us figure out what went wrong!

instead of

{
  "role": "tool_calls",
  "time": 1741604787317,
  "message": "",
  "toolCalls": [
    {
      "id": "call_2UfxMFreci5XJvIcFVc8ovVR",
      "type": "function",
      "function": {
        "name": "confirmReservationWithEmail",
        "arguments": "{\n  \"date\": \"2025-03-10\",\n  \"time\": \"14:30\",\n  \"email\": \"antoine.marcel@hotmail.fr\",\n  \"lastname\": \"Marcel\",\n  \"firstname\": \"Antoine\",\n  \"people_count\": 2,\n  \"phone_number\": \"\"\n}"
      }
    }
  ],
  "secondsFromStart": 41.882
}
{
  "role": "bot",
  "time": 1741604788313,
  "source": "",
  "endTime": 1741604791453,
  "message": "C'est bon Antoine votre table pour deux personnes est réservée aujourd'hui",
  "duration": 3140,
  "secondsFromStart": 46.34
}
{
  "name": "confirmReservationWithEmail",
  "role": "tool_call_result",
  "time": 1741604789399,
  "result": {
    "result": "Réservation confirmée"
  },
  "toolCallId": "call_2UfxMFreci5XJvIcFVc8ovVR",
  "secondsFromStart": 43.964
}

Try setting the

async

parameter to

false

in your tool configuration to ensure that the AI waits for the

tool_call_result

before responding. This will make your tool calls synchronous and should prevent premature responses. If this doesn’t help, you might want to confirm your server is returning responses correctly, as per the VAPI documentation.