Add get_sandbox_request hook for per-rollout customization

[rasdani]

Adds a get_sandbox_request(state) method that subclasses can override to customize the sandbox request per-rollout.

Use case: Environments with dynamic docker images per example (e.g., SWE-bench has different images per task).

Change:

• get_sandbox_request() returns self.sandbox_request.model_copy()
• setup_state() uses self.get_sandbox_request(state) instead of self.sandbox_request

Backwards compatible - existing subclasses work unchanged.

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

Enables per-rollout customization of sandbox creation without altering existing subclasses.

• Add get_sandbox_request(state) in sandbox_env.py returning self.sandbox_request.model_copy() (overrideable)
• Update SandboxEnv.setup_state() and RLMEnv._recreate_sandbox() to call get_sandbox_request(state) when creating sandboxes
• Keeps existing behavior by default via a copied base request

^{Written by Cursor Bugbot for commit a9523de. This will update automatically on new commits. Configure here.}

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ rasdani
❌ cursoragent
_{You have signed the CLA already but the status is still pending? Let us recheck it.}