Inconsistent, unexplainable "Internal Server Error" in Google Playbook

Google Cloud Build with AI Conversational Agents (Dialogflow CX)
1

Hi,

We seem to be getting random “Internal Server Error” calls in Playbooks. The playbook is very simple and calls a tool to get answers from a custom RAG solutiion (called via a cloud run function). The tool seems to start running but the Playbook instantly errors with the “Internal Server Error”.

What can I do to help debug this? There is very little in the way of debugging information from the Playbook itself. Anything we should check for?

Thanks!

2

Hi @mark_jones_ebm,

This issue might be related to response timeout in Dialogflow CX. Dialogflow expects a webhook response within your configured timeout (up to 30 seconds). If the Cloud Run Function experiences a cold start, it may cause a delay that makes the dialogflow webhook to timeout.

You can try setting a minimum number of instances in Cloud Run or increasing the webhook timeout in the Dialogflow CX console.

It is also best to check your Cloud Logging and Cloud Run Logs for more details about the error. If this doesn’t resolve the issue, it’s best to contact Google Cloud Support with the detailed information.

3

Hi @caryna,

I am not using CX here but Playbooks calling tools (which as far as I can see, do not have a timeout setting). Also, the error sometimes occurs as soon as the tool is called.

I will raise with support.

Thanks

4

Hey @mark_jones_ebm ,

I had the same issues connecting my database lookup API and it is really frustrating to find out what is the issue.

For me it would usually boil down to:

  • An actual server error (I dont think applicable to you) ~5% of the time
  • A small issue in the yaml openapi spec, openapi version stuff, dumb yaml syntax stuff… (could be your issue) ~most of the time

Ps: I would sometimes use the browser’s inspect tool and it sometimes showed a console error to elucidate, even the network section of your browser’s dev tools could help.

5

Thanks for that. The strange this was when it did error, it instantly errored. It’s calling a cloud run function so they are rarely up and down.

The issue seems to have solved itself so i would imagine it was a Playbooks bug that has now been fixed. That’s what I am hoping for anyway!