feat: use EUC for AI generation functions when connection_id is not present
Fixes #460856043 🦕
doctest failures appear to be related
FAILED bigframes/bigquery/_operations/ai.py::bigframes.bigquery._operations.ai.generate FAILED bigframes/bigquery/_operations/ai.py::bigframes.bigquery._operations.ai.generate_bool FAILED bigframes/bigquery/_operations/ai.py::bigframes.bigquery._operations.ai.generate_double FAILED bigframes/bigquery/_operations/ai.py::bigframes.bigquery._operations.ai.generate_int
Yeah those look like permission failures. I guess our doc runners don't have EUC permissions for AI functions.
cl/828515544 may be a useful example. that's where I granted the samples tests users permissions. that shouldn't affect doctest, though. doctest should use the compute engine service account from bigframes-dev.
cl/828515544 may be a useful example. that's where I granted the samples tests users permissions. that shouldn't affect doctest, though. doctest should use the compute engine service account from bigframes-dev.
Would doctest be considered a background job (I'm not very familiar with these setups)? If that's true, then we will need to explicitly provide a connection_id too 🧐.
Do not merge this PR until related failures in doc tests and notebook tests are gone
Check out this pull request on ![]()
See visual diffs & provide feedback on Jupyter Notebooks.
Powered by ReviewNB
e2e failures appear unrelated
FAILED tests/system/large/functions/test_remote_function.py::test_remote_function_max_instances[set-None]
FAILED tests/system/large/functions/test_remote_function.py::test_remote_function_max_instances[no-set]
FAILED tests/system/large/streaming/test_bigtable.py::test_streaming_df_to_bigtable
FAILED tests/system/large/streaming/test_pubsub.py::test_streaming_df_to_pubsub