The Scenario
A model governance review is coming up Thursday. Your ML platform lead handed off a dataset — my-project.ml_models — and asked you to build a model inventory before the meeting: model names, types, training dates, evaluation metrics, hyperparameters. All twelve models. In a shared Google Sheet.
She left for a conference Wednesday morning. The handoff document says "you can pull this from BigQuery." It does not say how.
The bad version:
- Search BigQuery documentation for the INFORMATION_SCHEMA tables that expose ML model metadata. Find that BigQuery ML metadata lives in BQML_MODELS and BQML_TRAINING_INFO views.
- Write a query that joins those views for the right project and dataset, run it in the BigQuery console, check the output, realize the hyperparameter columns come back as JSON structs that don't paste cleanly into a sheet.
- Manually expand each struct into rows. Do this twelve times.
This is not analysis work. It is data extraction, reformatting, and entry — and Thursday is not very far away.
The Easy Way: One Prompt in SheetXAI
SheetXAI is an AI agent inside your Google Sheet. It reads your sheet and talks to Google BigQuery directly — including querying BigQuery ML model metadata and writing it into structured columns.
List all BigQuery ML models in my dataset my-project.ml_models and for each one write the model name, model type, training date, and evaluation metrics into my Google Sheet
What You Get
- One row per model, with model name, model type (LOGISTIC_REG, BOOSTED_TREE_CLASSIFIER, etc.), training completion date, and the evaluation metrics BigQuery exposes for that model type.
- Metrics land as separate columns — not as a JSON blob — so you can sort, filter, or format them without further processing.
- If a model has no evaluation data yet (still training, or no eval run), that's noted in the relevant cell rather than crashing the output.
What If the Data Is Not Quite Ready
You also need hyperparameter values per model
Fetch details for all BigQuery ML models in my dataset my-project.ml_models and export their names, training loss, evaluation loss, and hyperparameters into my Google Sheet — one column per hyperparameter type, one row per model.
You want to filter to only classification models
List all BigQuery ML models in my-project.ml_models where the model type is a classification model (LOGISTIC_REG or BOOSTED_TREE_CLASSIFIER), and write the model name, type, training date, and AUC-ROC metric into my Google Sheet.
Add a column flagging models that haven't been retrained in 90 days
List all BigQuery ML models in my-project.ml_models, write name, type, training date, and evaluation metrics into my Google Sheet, and add a column called 'retrain_flag' — mark 'overdue' for any model whose training date is more than 90 days ago, otherwise leave it blank.
Pull the inventory, rank models by evaluation loss, and note the best performer
Fetch all BigQuery ML models in my-project.ml_models, write their names, types, training dates, and evaluation loss into my Google Sheet sorted by evaluation loss ascending, and write 'best model' into a column called 'notes' for the row with the lowest evaluation loss.
Inventory, ranking, and annotation — one prompt, one pass.
Try It
Get the 7-day free trial of SheetXAI and open a Google Sheet, then ask it to pull your BigQuery ML model inventory before your next governance review. You can also audit the full dataset schema or return to the Google BigQuery integration overview.