You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm using the BBC fulltext table from the public BQ for demonstration purposes here, I copied it to a private dataset that I own directely on BQ. The agent's query on the logs was successful and it retrieved the correct row, but either the results weren't passed correctly or LLM didn't understand how to retrieve the query results, resulting in a failed query fallback. The same problem also occurs in private datasets that I have.
Environment
Windows 11
Virtual Python Environment 3.12, configured according Data Science ReadMe instructions.
Reproduction steps or code
1- Download data-science module
2- Build the environment
3- Copy the Fulltext table to a private dataset e.g
4- Run The Agent
Error log
Translated Question: "I want the first fulltext that contains Microsoft in the title"
Important Logs:
The Query Below Works normally if executed directely on BQ:
And also works on the tool:
The next 2 screenshots are the same event:
The next 2 screenshots are the same event too:
Here we can notice that despite the right row was obtained before, the agent cannot return the answer to me