Vertex AI Agent Builder- BigQuery chat conversation export and evaluation pipeline

[DevBitcot]

Hello,

I'm looking to set up an evaluation pipeline for my Vertex AI agent application and am in search of any helpful documentation or tutorial videos that might be available. Additionally, I've configured the export of conversation history to a BigQuery table, but it appears that no data is being captured. If anyone has experience with these issues or can offer any assistance, it would be greatly appreciated. Thank you!

@xavidop

[xavidop]

Hi, did you follow all the steps here: https://cloud.google.com/dialogflow/cx/docs/concept/export-bq

did you interact with the agent, it only adds rows if you interact with it. Do you have bigquery APIs enabled in the same GCP project?

[DevBitcot]

Yes, I have completed all these steps except for interaction logging, as I didn't find an option to enable it in my Vertex agent. However, I have activated cloud logging, conversation history, and BigQuery export, including specifying my dataset and table name. I've also added the schema. Additionally, I've interacted with the agent, which has a significant amount of data for export, and I've enabled the API in the same project.

[xavidop]

Hi, you have to enable the interaction logging to have the export to BQ working: https://cloud.google.com/dialogflow/cx/docs/concept/agent#settings-general

[DevBitcot]

these are the options i have enabled. i didn't see any other options like interaction logging.

[xavidop · Accepted solution]

Thats odd, I cannot find it either! did they remove it? Please create a ticket here: https://issuetracker.google.com/issues/new?component=956347

[DevBitcot]

Thanks for your help.
Raised the ticket.
this is another ticket (https://issuetracker.google.com/issues/343035071) which i raised a week ago, still unresolved. Please have a look.