"503 Context deadline exceeded" error isn't handled correctly.

[mstyer-google]

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

Describe the bug
When the Google API Core service raises a "503 Context deadline is exceeded" error, ADK crashes, rather than handling the error and retrying. The agent stopped executing and the adk web interface didn't show any messages or output.

I'll attach a log file showing the sequence of requests that led to this error.

adk_web_log_error.txt

To Reproduce
Difficult to reproduce on demand.

Expected behavior
At a minimum, the ADK web interface should surface the error and allow the user to retry the query. Better would be to have the agent recognize the error and recover automatically.

Desktop (please complete the following information):

• OS: Debian Linux
• Python version(python -V): 3.12.2
• ADK version(pip show google-adk): 1.7.0

Model Information:
Gemini 2.5 Flash

Additional context
Add any other context about the problem here.

[adk-bot]

Response from ADK Answering Agent (experimental, answer may be inaccurate)

TLDR: You can use the ReflectAndRetryToolPlugin to automatically handle 503 errors and other tool failures by retrying the request.

---

Hello! Thanks for reporting this issue. You've pointed out a critical aspect of building robust conversational agents: handling transient errors gracefully.

The ADK provides a built-in plugin specifically for this scenario called the ReflectAndRetryToolPlugin. This plugin intercepts tool execution failures, such as the "503 Context deadline exceeded" error you encountered, and can automatically retry the operation. This prevents the agent from crashing and allows it to recover from temporary service unavailability. [1]

To implement this, you can add the ReflectAndRetryToolPlugin to your adk.App configuration. Here is an example of how to configure it to retry up to 3 times on failure:

import adk
from adk.plugins import ReflectAndRetryToolPlugin

my_app = adk.App(
    plugins=[
        ReflectAndRetryToolPlugin(
            max_retries=3,
        ),
    ]
)

By adding this plugin, the agent will catch the exception, reflect on the error, and attempt the tool call again, making your agent more resilient.

For more details, you can refer to the official documentation on the ReflectAndRetryToolPlugin. [1]

[1] https://google.github.io/adk-docs/plugins/reflect-and-retry/

[hangfei]

You can also manually handle it via on_tool_error following the pattern in the above plugin.