When I'm using Gemini, I encounter issues when I make a tool call. I have no issues with Groq or OpenAI using the same code. (Using their OpenAI compatibility URLs)

Gemini 2.0 Flash enforces stricter payload requirements than Groq or OpenAI. For tool calls, ensure that STRING parameters use only the 'enum' format and that your response is a stringified JSON including the tool call ID, with no line breaks. For example:

json
{
  "parameters": {
    "insuredEmail": "user@example.com" // must be an enum value if STRING type
  },
  "tool_call_id": "abc123"
}

Double-check your payload structure to match Gemini's expectations to avoid errors like "Invalid JSON payload received..." [documentation](https://discord.com/channels/1211482211119796234/1344710846865084616). Sources: - [Gemini 2.0 Flash tool call implementation](https://discord.com/channels/1211482211119796234/1344710846865084616) - [Gemini 2.0 Flash tool call troubleshooting](https://discord.com/channels/1211482211119796234/1332510949940133960)

I don't understand what this documentation is saying

Hey, Could you please schedule a brief call at your convenience so we can discuss this matter in detail? Kindly use the following link to book a suitable time: .

Bro the next available call is june

We recommend you to use OpenAI or Anthropic for more stable toolCalls

So does that mean gemini tool calling is not compatible with Vapi?

For latency issues, can you send some call ids so we can look into it?

Its not vapi latency side issues its simply OpenAI have slow TTFT

sg.