google
`adk web` does not render Eval tab when more than one agent folder exists
Describe the bug
The Eval tab in adk web UI is blank if there is more than one agent/folder present.
To Reproduce
Generate minimal boilerplate for a new single-agent system:
mkdir -p ~/project/test_01 && cd $_
python -m venv .venv
source .venv/bin/activate
pip install google-adk
pip list | grep google-adk # 1.19.0
adk create my_agent --model gemini-2.5-flash --project ${GOOGLE_CLOUD_PROJECT:?} --region us-central1
Start the web server:
adk web
Click the link to open the web browser.
Switch to Eval tab. The tab is not empty. There are UI elements to create a new evaluation set and some helpful text. So far so good!
Close the browser tab. Terminate the web server.
Now let's reproduce the issue...
Create boilerplate for the second agent:
adk create your_agent --model gemini-2.5-flash --project ${GOOGLE_CLOUD_PROJECT:?} --region us-central1
Start the web server again:
adk web
Click the link to open the web browser. Verify that now there are two agents:
It does not matter which one you choose, I'll go with my_agent here.
Once the agent is chosen, switch to the Eval tab. The tab is empty!
This is a bug, I believe!
Close the browser tab. Terminate the web server.
To verify that the existence of the second agent is an issue, delete the second agent:
rm -rf your_agent
Start the web server:
adk web
Click on the link to open a browser tab. The only available agent is pre-selected:
Switch to the Eval tab. The UI is back!
I have observed this behaviour since I updated ADK from 1.18 to 1.19.
All other interactions with the agents remain unaffected, no other issues.
Expected behavior The Eval tab UI should be available and work for multi-agent systems, as before.
Screenshots Included.
Desktop (please complete the following information):
- OS: Windows 11
- Browser: Google Chrome, latest updates installed as of time of writing
- Python version(python -V):
3.12.3on Linux - ADK version(pip show google-adk):
1.19.0
Model Information:
- Are you using LiteLLM: No
- Which model is being used: irrelevan
Another thing I have noticed this issue we can't able to see the evaluation results in the UI, if we click the pass or fail we should able to see the evaluation run result in UI with pass / fail indication and on hover the chat message we could able to see the expected result and actual result these are not showing all the time means I can able to see the indication only few time most of the time the indication is show (not rendering the evaluation results with the pass/fail indication)
Another thing I have noticed this issue we can't able to see the evaluation results in the UI, if we click the pass or fail we should able to see the evaluation run result in UI with pass / fail indication and on hover the chat message we could able to see the expected result and actual result these are not showing all the time means I can able to see the indication only few time most of the time the indication is show (not rendering the evaluation results with the pass/fail indication)
I can confirm that I had experienced this as well! Although for me this started before 1.19.0 while the current issue has started with 1.19.0.