Checkpoints¶
Truthound checkpoints are the bridge between local validation and operational workflows. Use this page as the entry point for CI/CD runs, scheduled validation, routing, escalation, notifications, and production-oriented checkpoint design.
Who This Guide Is For¶
Use the checkpoint docs when you need to:
- run repeatable validation pipelines in CI or scheduled jobs
- connect validation outcomes to routing, throttling, escalation, and notifications
- publish docs or persist results from checkpoint execution
- understand how the checkpoint layer relates to the 3.0 kernel and result model
The Checkpoint Mental Model¶
Treat checkpoints as orchestration over the kernel, not as a second validation engine.
- validation still resolves to the canonical
ValidationRunResult - checkpoint metadata adds automation context such as run status, actions, and routing state
- persistence, notifications, CI summaries, and docs updates should consume
validation_runorvalidation_view, not reconstruct a parallel result model
Start Here¶
| Task | Best page |
|---|---|
| Understand the checkpoint system end to end | Checkpoint Family Overview |
| Build your first checkpoint | Basics |
| Add scheduled or event-driven execution | Triggers |
| Route issues to different teams or workflows | Routing |
| Control notification volume or duplicate runs | Throttling and Deduplication |
| Add multi-step operational handling | Escalation |
| Understand async or higher-throughput execution | Async Execution |
| Integrate with CI vendors | CI Platforms and CLI checkpoint commands |
Recommended Reading Path¶
Runtime Model¶
CheckpointResult.validation_runholds the canonicalValidationRunResultCheckpointResult.validation_viewis the compatibility summary surface for legacy formatting pathsCheckpointResult.validation_resultis removed in 3.0
Operational Boundaries¶
Persistence boundary¶
Storage-facing DTO conversion is allowed at the persistence edge, but checkpoint orchestration should remain centered on canonical runtime results.
Reporting boundary¶
Validation docs and reporters should build from ValidationRunResult directly rather than inventing a second report model.
Automation boundary¶
Routing, throttling, escalation, and notifications should work from checkpoint metadata plus validation_run/validation_view.