# Constal Evals documentation

> Evals concepts, workflows, operations, and reference for humans and agents.

- [Evals](https://constal.ai/docs/evals.md): Measure Agent behavior with versioned cases and Scorers, prevent regressions, and turn proven data into trained Models.
- [Evals SDK types and validators](https://constal.ai/docs/evals/sdk.md): Validate Scorers, Suite inputs, and training recipes in code before sending them through the Platform API.
- [Evals with the CLI](https://constal.ai/docs/evals/cli.md): Create Datasets and Scorers, operate suites and baselines, and automate metered SFT or RL jobs.
- [Create and publish Datasets](https://constal.ai/docs/evals/datasets.md): Turn representative examples or recorded Agent interactions into immutable, reproducible Dataset versions.
- [Create and calibrate Scorers](https://constal.ai/docs/evals/scorers.md): Define reusable code, Judge, or human judgments for suites, gates, validation, and reinforcement-learning rewards.
- [Run suites and compare results](https://constal.ai/docs/evals/suites.md): Evaluate one pinned Agent across an immutable Dataset, then set baselines and detect regressions.
- [Run continuous evaluations](https://constal.ai/docs/evals/schedules.md): Repeat one pinned Suite on a durable cadence without overlapping work or creating catch-up storms.
- [Train with SFT or reinforcement learning](https://constal.ai/docs/evals/training.md): Run durable training through an exact provider, inspect content-addressed checkpoints, and create a candidate Model through a compatible Gateway.
- [Build a Training Provider](https://constal.ai/docs/evals/training-providers.md): Package a training system as a provider-neutral Connection with exact model discovery, durable reconciliation, and content-addressed checkpoints.