Outputs builder

[willccbb]

Description

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Test improvement

Testing

• All existing tests pass when running uv run pytest locally.
• New tests have been added to cover the changes

Checklist

• My code follows the style guidelines of this project as outlined in AGENTS.md
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Any dependent changes have been merged and published

Additional Notes

---

Note

Shifts generation outputs from raw State objects to serialized RolloutOutput dicts and introduces an incremental results builder for efficient saving and display.

• Add RolloutOutput type and update GenerateOutputs to use outputs (replacing states); update docs and all call sites/tests accordingly
• Introduce GenerateOutputsBuilder to serialize once, stream progress, compute metadata, and support intermediate saves; final results sorted by example_id
• Add state_to_output/states_to_outputs, remove sanitize_states; validate state_columns are JSON-serializable; errors serialized as strings
• Update Environment.generate() to build outputs incrementally and adjust progress callbacks to use RolloutOutput
• Modify utils (eval_utils, eval_display, logging_utils, save_utils) to consume serialized outputs; make_dataset now builds directly from outputs
• Adapt integrations (gepa.adapter, RL trainer orchestrator) to the new outputs schema
• Update Gym/eval/CLI tests and fixtures to the new API

^{Written by Cursor Bugbot for commit a17a1d1. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

[mikasenghaas]

lfgtm!

[mikasenghaas]

ah yea funnily enough i had a type very similar to this one initially but then scratched it bc it felt too redundant. basiclly RolloutOutput = State - untyped columns + state columns and so I wasn't sure this justified an entirely new type (that now also has to be synced)