Run Artifacts And Observability

Summary

Run artifacts and structured results make automation explainable after the work has finished.

Problem

Logs alone are weak operational evidence. They can be noisy, transient, and hard to connect to a specific business outcome. Operators need counts, payload summaries, errors, progress, and artifacts tied to run identity.

Principle

Emit structured run data. Keep progress and result payloads low-cardinality, auditable, and useful to UI, diagnostics, and future agents.

Platform Shape

The execution modes doc describes ops.job_run.progress_json, structured Cloud Tasks dispatch logs, and stale run recovery. Execution-state services own job diagnostics and read models, while observability support lives under services/ops/observability.

Source Evidence

• docs/reference/job-execution-modes.md
• services/ops/execution_state/jobs
• services/ops/observability

Related Reading

• docs/engineering/reference/ops-diagnostics.md
• docs/engineering/concepts/job-run-state.md