Course overview/Operate with an agent3 of 5
Logs, history, and the bill
Logs, history, and the bill
Use structured run evidence for triage and distinguish local logs from warehouse-side cost history.
Leave evidence behind
Bruin stores run evidence under logs/runs/<pipeline>/<run-id>.json, relative to the Git project root - the directory containing .bruin.yml. bruin validate pipeline --output json is useful in CI. In Bruin Cloud, an agent can read past runs, the status of each table in a run, logs, and validation errors through read-only tools such as pipeline-run-list, asset-instance-list, asset-instance-logs, asset-runs, and validation-error-list.
DuckDB has no equivalent billing or shared query-history service. BigQuery and Snowflake expose query text, user, permissions, timing, bytes or credits, and cost through warehouse-side history. Treat cost claims as platform-specific evidence.
Your task
Write docs/observability.md describing the local log path, a JSON validation command, the Bruin Cloud read-only tools, and the BigQuery or Snowflake evidence you would inspect. Include one read-only triage query command with --description.
Check your understanding
- What belongs in a run log?
- Why is read-only run-history access useful to an agent?
- Does local DuckDB provide a warehouse bill?
Do it with your agent
Say next lesson, gather the evidence and write the observability note, then say review my work.
Rubric
- Includes
logs/runs/<pipeline>/<run-id>.jsonandbruin validate pipeline --output json. - Names run history, the status of each table in a run, logs, and validation errors as Cloud read-only information.
- Names the read-only Cloud tools that expose that information.
- Distinguishes DuckDB from BigQuery and Snowflake query history and cost evidence.