feat(swe_env): decoupled SWE environment library + verifier (#1249) [1/3]

[adil-a]

Decoupled SWE environment library + verifier (NVIDIA-NeMo#1249)

Stacked PR 1 of 3 · base: main · next: stack/2-swe-agents-openhands

Part of #1249 — decouple the SWE environment infrastructure from the agent harnesses — built on top of the Sandbox API from NVIDIA-NeMo#1377 (now on main).

What this PR adds

Today the entire SWE-bench stack (sandbox provisioning, the per-dataset eval recipes, patch parsing, and grading) is fused into the OpenHands agent in responses_api_agents/swe_agents/app.py. No other agent can reuse any of it. This PR extracts that infrastructure into two reusable, agent-agnostic pieces. It adds the new pieces only — no existing agent is rewired yet (that is PR 2).

1. responses_api_agents/swe_env/ — a shared SWE environment library. An installable distribution that any agent imports to provision and drive its own task sandbox on top of the NVIDIA-NeMo#1377 Sandbox API:

• environment.py — AsyncSweEnvironment, the provider-neutral execute/upload/download wrapper agents drive.
• harness.py / registry.py — the SweTaskHarness ABC and a name→harness registry; one source of truth for the dataset families (swebench, swe-rebench, nv-internal, swe-bench-ext, r2egym, flat-eval).
• grading.py — pure compute_resolved() / reward_from_report() helpers.
• parsing/ — the SWE-bench-Ext test-log parser (relocated here as the single home; the duplicate under swe_agents is removed in PR 2).
• providers/{apptainer,docker}_provider.py — SandboxProvider implementations, including the Apptainer .sif path the on-prem clusters use.
• lifecycle.py — a thin acquire_sandbox() context manager. (The heavier durable-registry/reaper/admission lifecycle layer is intentionally out of scope here and tracked as a follow-up to generalize it into the sandbox layer.)
• model_endpoint.py — provider-neutral resolution of a sandbox-reachable model-server endpoint for self-driving agents.

2. resources_servers/swe_env/ — the verifier. A SimpleResourcesServer whose verify() is the single scoring entry point: it extracts the model's patch from the response, grades it in a fresh, stateless sandbox, and returns the eval-side fields. verify_task.py holds the server-private grading orchestrator (reset → apply patch → run eval → parse → reward). Ships with a launchable config + offline example fixtures for the data gate.

Tests

responses_api_agents/swe_env/tests/   168 passed, 1 skipped
resources_servers/swe_env/tests/        14 passed, 3 skipped

ruff check / ruff format clean. Real-container grading paths are skip-guarded (apptainer/Docker); the orchestrator is exercised via a FakeSandbox that replays recorded eval transcripts.

Review notes

• Pure additions: 59 new files, no changes to existing code — safe to review in isolation.
• The verifier imports the library as a packaged distribution (added to its requirements.txt), so the cross-tree import works under ng_run/ng_test.

🤖 Generated with Claude Code

[adil-a]

Superseded by the stacked PRs opened directly on NVIDIA-NeMo/Gym (NVIDIA-NeMo#1677, NVIDIA-NeMo#1678, NVIDIA-NeMo#1679).