Incorrect Role Assignment in Multi-Agent Mode Causes Context Misinterpretation

[rick-bm]

** Please make sure you read the contribution guide and file the issues in the right place. **

Describe the bug
In multi-agent mode, the role of the agent's response in the context is incorrectly marked as "user." This causes downstream agents to mistakenly treat the previous agent's reply as user input, leading to incorrect analysis.

To Reproduce
Steps to reproduce the behavior:

1. Set up a multi-agent environment.
2. Allow one agent to generate a response.
3. Pass the context to a downstream agent.
4. Observe that the downstream agent treats the previous agent's response as user input, which leads to errors.
   e.g. https://adk.wiki/agents/custom-agents/#full-code-example

Expected behavior
The role of each agent's response should be correctly marked to ensure downstream agents interpret the context accurately. Specifically, agent responses should not be marked as "user."

Screenshots

Desktop (please complete the following information):

• OS: [e.g. iOS] MacOs
• Python version(python -V): 3.13.3
• ADK version(pip show google-adk): v1.2.1

Model Information:
For example, which model is being used.

Additional context
Add any other context about the problem here.

[coder-aditi]

Hello @hangfei , I added the changes to the code to fix this issue, kindly review and let me know your suggestions or if any changes required

[wuliang229]

@rick-bm this is working-as-intended. In multi-agent system, if there is a response by Agent 1 followed by Agent 2, then the response by Agent 1 is converted to have "role": "user". See the implementation here