Multiple function_response in single API call causes "Proto field is not repeating" error

[lina-bmx]

Describe the bug
When the Gemini model returns multiple function calls in a single
response, the ADK fails to send the combined function responses back
to the API. The Gemini API rejects the payload with a protobuf
serialization error: "Proto field is not repeating, cannot start
list."

To Reproduce
Steps to reproduce the behavior:

1. Install @google/adk@0.2.4
2. Create an agent with multiple FunctionTool tools
3. Send a prompt that causes Gemini to call 2+ tools simultaneously
4. All tools execute successfully and return results
5. See error:

  400 Invalid JSON payload received. Unknown name "response" at        
  'contents[2].parts[0].function_response': Proto field is not         
  repeating, cannot start list.        

Expected behavior
The ADK should properly format multiple function responses in a way
the Gemini API accepts, allowing the model to see all tool results
and generate a combined response.

Screenshots
N/A - error appears in console logs

Desktop (please complete the following information):

• OS: Windows 10
• TS version/environment: TypeScript 5.x, Node.js v22.20.0
• ADK version: @google/adk@0.2.4

Additional context

• Single tool calls work fine - issue only occurs with 2+
  simultaneous tool responses
• Tools execute successfully; the failure is in sending responses
  back to Gemini
• Model: gemini-3-flash-preview, Backend: VERTEX_AI
• Related Python ADK issue:
  Add enable_parallel_tool_execution flag to LlmAgent adk-python#1076
• Workaround: Catch error and return fallback message, but model
  never sees tool results

[Varun-S10]

Hi @lina-bmx , we tried to reproduce the issue, but it is working fine on our end. We were able to call three function tools in one prompt without any errors. Could you please share a reproducible code sample so we can investigate this further?

import { FunctionTool, LlmAgent } from '@google/adk';
import { z } from 'zod';

// Tool 1: Get Weather
const getWeather = new FunctionTool({
  name: 'get_weather',
  description: 'Returns the current weather for a city.',
  parameters: z.object({
    city: z.string().describe('The city name'),
  }),
  execute: async ({ city }) => {
    console.log(`[Tool] Executing getWeather for ${city}...`);
    return { weather: 'Sunny', temperature: '22°C' };
  },
});

// Tool 2: Get Time
const getTime = new FunctionTool({
  name: 'get_time',
  description: 'Returns the current time for a city.',
  parameters: z.object({
    city: z.string().describe('The city name'),
  }),
  execute: async ({ city }) => {
    console.log(`[Tool] Executing getTime for ${city}...`);
    return { time: '10:00 AM', timezone: 'GMT' };
  },
});

// Tool 3: Get Date
const getDate = new FunctionTool({
  name: 'get_date',
  description: 'Returns the current date for a city.',
  parameters: z.object({
    city: z.string().describe('The city name'),
  }),
  execute: async ({ city }) => {
    console.log(`[Tool] Executing getDate for ${city}...`);
    return { date: '10 Oct', year: '2025' };
  },
});

// Create the Agent
export const rootAgent = new LlmAgent({
  name: 'multi_tool_tester',
  model: 'gemini-3-flash-preview', // As requested
  description: 'An agent to test simultaneous tool calls.',
  instruction: 'You are a helpful assistant. Use tools to answer questions about date, time and weather.',
  tools: [getWeather, getTime, getDate],
});

We have added the code. Please try it out and let us know the results.