verifiers 0.2.2.dev40__tar.gz → 0.2.2.dev41__tar.gz

{verifiers-0.2.2.dev40 → verifiers-0.2.2.dev41}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: verifiers
-Version: 0.2.2.dev40
+Version: 0.2.2.dev41
 Summary: Verifiers: Environments for LLM Reinforcement Learning
 Project-URL: Homepage, https://github.com/primeintellect-ai/verifiers
 Project-URL: Documentation, https://github.com/primeintellect-ai/verifiers

{verifiers-0.2.2.dev40 → verifiers-0.2.2.dev41}/verifiers/v1/harness.py

@@ -9,7 +9,7 @@ from typing import TYPE_CHECKING, ClassVar, Generic, TypeVar
 from verifiers.v1.clients import ModelContext
 from verifiers.v1.configs.harness import HarnessConfig
 from verifiers.v1.decorators import discover_decorated, invoke_all
-from verifiers.v1.errors import HarnessError, boundary
+from verifiers.v1.errors import HarnessError, SandboxError, boundary
 from verifiers.v1.runtimes import ProgramResult, Runtime
 from verifiers.v1.task import TaskData
 from verifiers.v1.types import Messages
@@ -140,6 +140,11 @@ class Harness(ABC, Generic[ConfigT]):
         if result.exit_code != 0:
             # The real cause is at the END of a traceback, so keep the tail.
             detail = (result.stderr or result.stdout).strip()[-2000:] or "<no output>"
+            if not await runtime.alive():
+                raise SandboxError(
+                    f"runtime died under harness {self.config.id!r} "
+                    f"(exit {result.exit_code}): {detail}"
+                )
             raise HarnessError(
                 f"harness {self.config.id!r} exited {result.exit_code}: {detail}"
             )

{verifiers-0.2.2.dev40 → verifiers-0.2.2.dev41}/verifiers/v1/runtimes/base.py

@@ -197,6 +197,16 @@ class Runtime(ABC):
     async def run(self, argv: list[str], env: dict[str, str]) -> ProgramResult:
         pass
 
+    async def alive(self) -> bool:
+        """Whether the box still executes anything. Not every runtime raises when
+        the box is gone — some surface it as `exec`'s own non-zero result,
+        indistinguishable from the command failing. One probe on the failure path
+        tells the two apart before we blame anyone."""
+        try:
+            return (await self.run(["true"], {})).exit_code == 0
+        except Exception:  # noqa: BLE001 - failing to exec at all means the box is gone
+            return False
+
     async def run_program(self, argv: list[str], env: dict[str, str]) -> ProgramResult:
         """Run the harness's MAIN program — the rollout itself (a possibly long-lived, stateful,
         agentic run) — as opposed to the short idempotent infra ops (write / mv / install /