Daily BigQuery slow-query regression hunter to GitHub tuning issue

This workflow finds queries that got slower or more expensive overnight and files an actionable tuning ticket so the regression doesn't quietly drain your BigQuery budget. Instead of a generic alert, the issue carries the offending query text, its execution-stage plan, and a before/after cost delta so an engineer can start tuning immediately.

Run it on data platforms where scheduled queries and dbt models change often and a single bad join or dropped partition filter can 10x slan run's cost. Best for teams who want a paper trail of regressions in GitHub rather than a noisy Slack feed.

1. 1A daily schedule fires after the prior day's jobs have settled.
2. 2BigQuery's INFORMATION_SCHEMA.JOBS is queried for completed jobs and their total_bytes_billed and runtime.
3. 3The same query hashes are pulled from a 14-day baseline to compute per-query deltas.
4. 4A logic step keeps only queries that regressed beyond the runtime or cost threshold.
5. 5For each survivor, the query plan and cost delta are formatted into a tuning report.
6. 6A GitHub issue is opened with the plan, deltas, and a 'tuning' label as the final output.