Empty response after tool call

Gemini API Documentation
,
1

Hi everyone, could you please help me?
I’m having an issue with my orchestrator agent using Google ADK.

I have a multi-agent system, where several subagents are being used as agent-as-tool.
For a specific query, I need a sequence of tools to be executed by different agents, one after another.

However, in several cases, after a subagent (used as tool) responds successfully, the root_agent receives the function response but then returns an empty text output and stops the reasoning process.

Here’s an example:

Response from agent-as-tool:

{
  "parts": [
    {
      "function_response": {
        "id": "8324785784357",
        "name": "genericagent",
        "response": {
          "blabla": [
            {
              "ihih": "eieie",
              "test": "hellouu"
            }
          ]
        }
      }
    }
  ],
  "role": "user"
}

Root agent output:

{
  "parts": [
    {
      "text": ""
    }
  ],
  "role": "model"
}

After this, the execution stops completely — as if the root agent lost its reasoning chain.
Has anyone faced something similar? How can I ensure the orchestrator continues reasoning after each tool response?

Thanks in advance!