Action Health Checks

Google monitors deployed actions to verify they're available and responding to user requests.

Each agent's endpoint is required to respond to a default, no parameter, "launch" intent. This supports the scenario where the user asks to talk to the agent without providing any specific need (e.g. "I want to talk to invocation name").

Common questions

Do I need to implement anything specific to support the health checks? No, agent health checks will function like a standard request to talk to an agent.

How often will my agent’s health be checked? Your agent may be monitored throughout the day.

How can I detect that a request is a health check? Google adds an argument to the request body so that you can detect a health check in your analytics. The launch intent request from Google will have an “is_health_check” argument passed to the intent, as the following example shows:

...
"arguments": [
  {
    "name": "is_health_check",
    "text_value": "1"
  }
]
...

What happens if Google detects that an agent is unhealthy? If Google detects that the agent is experiencing an outage, the agent and its actions are suspended. Users requesting the unhealthy agent will hear “agent is not available”.

How will I be notified if an agent is marked as unhealthy? Upon detecting an unhealthy agent, the account that was used to create or update the agent will receive an email from Google.

How can I check my agent's health status? You can check your developer console and see the status of your agents. Unhealthy agents will have a status of Unhealthy. Healthy agents will be marked as Deployed.

What’s the process for returning an agent to the deployed state? You should review your deployment and address any issues that might be causing the agent to fail to respond to a launch intent.

Google will continue to periodically monitor the agent's health while it is unhealthy. When Google detects that the agent is responding as expected, the agent will return to the deployed state.

How will I be notified if an agent is marked as healthy? Upon detecting an agent as healthy, the account that was used to create or update the agent will receive an email from Google.