verifiers 0.2.2.dev31__py3-none-any.whl → 0.2.2.dev33__py3-none-any.whl

verifiers/v1/__init__.py

@@ -37,7 +37,7 @@ from verifiers.v1.errors import (
 )
 from verifiers.v1.configs.harness import HarnessConfig
 from verifiers.v1.harness import Harness
-from verifiers.v1.configs.judge import JudgeConfig, JudgeSamplingConfig, Judges
+from verifiers.v1.configs.judge import JudgeConfig, Judges
 from verifiers.v1.judge import Judge, JudgeResponse, JudgeView
 from verifiers.v1.judges import (
     ReferenceJudge,
@@ -267,7 +267,6 @@ __all__ = [
     "Judge",
     "JudgeConfig",
     "Judges",
-    "JudgeSamplingConfig",
     "JudgeResponse",
     "JudgeView",
     "ReferenceJudge",

verifiers/v1/configs/judge.py

@@ -12,10 +12,6 @@ from verifiers.v1.types import ID, SamplingConfig
 from verifiers.v1.utils.install import env_name
 
 
-class JudgeSamplingConfig(SamplingConfig):
-    pass
-
-
 class JudgeConfig(BaseClientConfig):
     id: ID = ""
     """Plugin id; empty for a judge called directly by task code."""
@@ -23,7 +19,7 @@ class JudgeConfig(BaseClientConfig):
     """Reward key override for a plugged judge."""
     weight: float = 1.0
     model: str = "openai/gpt-5.4-nano"
-    sampling: JudgeSamplingConfig = JudgeSamplingConfig()
+    sampling: SamplingConfig = SamplingConfig()
     prompt: str | None = None
     prompt_file: Path | None = None
     """Prompt file override, mutually exclusive with `prompt`."""

verifiers/v1/dialects/responses.py

@@ -1,9 +1,9 @@
 """The OpenAI Responses dialect (codex and friends).
 
 Request parsing walks the `input` items, folding each run of assistant-side items (reasoning /
-assistant message / function_call) into one typed assistant message; response parsing reads the
-`output` items. Relay-only: the eval client forwards the program's bytes to a `/responses`
-endpoint and this dialect parses a copy for the trace. Server-side statefulness
+assistant message / function or custom tool call) into one typed assistant message; response
+parsing reads the `output` items. Relay-only: the eval client forwards the program's bytes to a
+`/responses` endpoint and this dialect parses a copy for the trace. Server-side statefulness
 (`previous_response_id`) is not emulated — the endpoint owns it.
 """
 
@@ -97,8 +97,7 @@ def parse_content(content) -> str | list[ContentPart]:
 
 
 def fold_assistant(items: list[dict]) -> AssistantMessage:
-    """One run of assistant-side items (reasoning / message / function_call) -> one typed
-    assistant message."""
+    """One run of assistant-side items -> one typed assistant message."""
     content = ""
     reasoning: list[str] = []
     calls: list[ToolCall] = []
@@ -106,12 +105,12 @@ def fold_assistant(items: list[dict]) -> AssistantMessage:
         if item.get("type") == "reasoning":
             reasoning += [s.get("text", "") for s in item.get("summary") or []]
             reasoning += [c.get("text", "") for c in item.get("content") or []]
-        elif item.get("type") == "function_call":
+        elif item.get("type") in ("function_call", "custom_tool_call"):
             calls.append(
                 ToolCall(
                     id=item.get("call_id", ""),
                     name=item.get("name", ""),
-                    arguments=item.get("arguments", ""),
+                    arguments=item.get("arguments", item.get("input", "")),
                 )
             )
         else:  # an assistant message item
@@ -151,12 +150,12 @@ def response_from_wire(response: OpenAIResponse) -> Response:
         elif kind == "reasoning":
             reasoning += [s.get("text", "") for s in item.get("summary") or []]
             reasoning += [c.get("text", "") for c in item.get("content") or []]
-        elif kind == "function_call":
+        elif kind in ("function_call", "custom_tool_call"):
             calls.append(
                 ToolCall(
                     id=item.get("call_id", ""),
                     name=item.get("name", ""),
-                    arguments=item.get("arguments", ""),
+                    arguments=item.get("arguments", item.get("input", "")),
                 )
             )
     tool_calls = calls or None
@@ -279,7 +278,10 @@ class ResponsesDialect(Dialect[dict, OpenAIResponse]):
                 run = []
             if assistant:
                 run.append(item)
-            elif item.get("type") == "function_call_output":
+            elif item.get("type") in (
+                "function_call_output",
+                "custom_tool_call_output",
+            ):
                 output = item.get("output")
                 content = (
                     parse_content(output)

verifiers/v1/gepa/adapter.py

@@ -87,7 +87,7 @@ class GEPAAdapter:
 
     def make_reflective_dataset(
         self,
-        candidate: Candidate,  # noqa: ARG002 - required by GEPA's adapter protocol
+        candidate: Candidate,  # Required by GEPA's adapter protocol.
         eval_batch: EvaluationBatch[Episode, Episode],
         components_to_update: list[str],
     ) -> Mapping[str, Sequence[Mapping[str, Any]]]:

verifiers/v1/utils/git.py

@@ -97,7 +97,7 @@ async def capture_patch(
             )
             return
         raw = await runtime.read(capped)
-    except Exception as exc:  # noqa: BLE001 - capture must never fail the rollout.
+    except Exception as exc:  # Capture must never fail the rollout.
         trace.info["patch_error"] = f"{type(exc).__name__}: {exc}"
         return
     finally:
@@ -105,7 +105,7 @@ async def capture_patch(
         # on shared-filesystem runtimes; removal is best-effort by design.
         try:
             await runtime.run(["rm", "-f", full, capped], env or {})
-        except Exception:  # noqa: BLE001,S110 - cleanup must never fail the rollout.
+        except Exception:  # Cleanup must never fail the rollout.
             pass
     if len(raw) > PATCH_CAP_BYTES:
         raw = raw[:PATCH_CAP_BYTES]

{verifiers-0.2.2.dev31.dist-info → verifiers-0.2.2.dev33.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: verifiers
-Version: 0.2.2.dev31
+Version: 0.2.2.dev33
 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.dev31.dist-info → verifiers-0.2.2.dev33.dist-info}/RECORD

@@ -167,7 +167,7 @@ verifiers/utils/threaded_sandbox_client.py,sha256=Pbr8MA4FDPEitL6z88S8T1qLJOmtXN
 verifiers/utils/tool_utils.py,sha256=gInWZODWQUZN4TyEMuuITyOUm9qlHJxtcvr1zZl_zJs,1020
 verifiers/utils/usage_utils.py,sha256=GPLC0xGY_Obrr8X7huWY-2iODZ7tgX-MtO8WSDN4rXI,3904
 verifiers/utils/version_utils.py,sha256=am3hZLnlaUWTFdllGZR1VMN91hEhiiE8FDmPf1cOG1k,2642
-verifiers/v1/__init__.py,sha256=z0yUgI1iVf8znsAtBG2fpRKuwipaQ5yS4PvGhWJbKtE,6859
+verifiers/v1/__init__.py,sha256=GiNNUcxzYkjZMnTXl5mKAlOSrxiaUkD1A-3U9rfiDso,6811
 verifiers/v1/agent.py,sha256=FP38GsZXuEJe3tJ-U2THpwBf9NWQeo847hWRZoeggNw,31956
 verifiers/v1/decorators.py,sha256=XRMkUQSyvXCYP5fOwzBYV5qEOlxLg6rzYhnhrHHHTIQ,3838
 verifiers/v1/env.py,sha256=w6sHWdWijLRZzhwVyJjadnGtziNqJAPCeMUoNdzHoLg,17712
@@ -217,7 +217,7 @@ verifiers/v1/configs/__init__.py,sha256=X7u6X7B3ieD1HZl79Tv1kg-yPHbT0QLvAebaGqxV
 verifiers/v1/configs/agent.py,sha256=N2Lm0iZXFxfsSIiXUQ7f3RDZcoin3M7pJj6HVdCE5eo,3537
 verifiers/v1/configs/env.py,sha256=2mEU1n2xG-v5PNdB2EJioAYP6Witnlny9kYi0jwKFQk,7286
 verifiers/v1/configs/harness.py,sha256=CG8hGbHl3rr6yawH3YYMAFl01M7Piiwf4w5oiw2Vagg,1564
-verifiers/v1/configs/judge.py,sha256=Y70YcLXX_m1NMYeO4Wl63t57NkHpK0XRjGYay2NonQ8,2575
+verifiers/v1/configs/judge.py,sha256=ZJlmWogHY8oDjvWvtNuVKvkzuheWalIjl1njJ4dE0Mk,2511
 verifiers/v1/configs/retries.py,sha256=1PY0hhPV2Q-KgBKKSUNBFccJzc_t1yG3P_44p8Q_1a4,771
 verifiers/v1/configs/task.py,sha256=L-j_CYIDP2JAF7CDA2FOZz8aQK2h7c0f1qnmJshWzu4,1069
 verifiers/v1/configs/taskset.py,sha256=IdZwsarn0_WvYdR_Kr4QYYdqbaoGYkbYyO5VA-RPYPc,657
@@ -233,7 +233,7 @@ verifiers/v1/dialects/__init__.py,sha256=PR3CJFI5siVJBXcF7H2Wi_UCygotHhL0Oj6DIOF
 verifiers/v1/dialects/anthropic.py,sha256=nj9J7OOhexmPHa-sWiNP2xlI9_bGhvwM5VWzdh19pBc,13526
 verifiers/v1/dialects/base.py,sha256=YZQDnasQseDXK-GHLfJLZgyHIIygS80YhD_sTjmUbfA,8126
 verifiers/v1/dialects/chat.py,sha256=DFvjmIW86jZUMWuz-hOqzxC6tiiFibiAzNEJ3JkpvjU,13998
-verifiers/v1/dialects/responses.py,sha256=vVgeT7-aWjtfED2IKlDwTATKn_AQZruHNL6mkFTTZTU,13548
+verifiers/v1/dialects/responses.py,sha256=ismnxUikPcPUIm80FUTPBbj-Cn1e0yGAsISMEseAWRc,13679
 verifiers/v1/envs/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 verifiers/v1/envs/agentic_judge/__init__.py,sha256=X7vQbbbfmJ_j-mJEfBkuB4a8QUPkUYkpG7yp0xYOEmg,279
 verifiers/v1/envs/agentic_judge/env.py,sha256=_zEKG9i3LsOVIQWULXYAHpbmrQNPWBRayJx2yjvtPbQ,15307
@@ -244,7 +244,7 @@ verifiers/v1/envs/single_agent/env.py,sha256=RuWYgdT2vVu0PBt1esmnlbCcYwUD6ZbmqT-
 verifiers/v1/envs/user_sim/__init__.py,sha256=nnT695HdqjAEouYL5wtVu6MhIp2wf5yuK-EYgxOVx0s,118
 verifiers/v1/envs/user_sim/env.py,sha256=u376-UJEVdzgcit7FXjz5KEwlPxSU3qqoIrxs6foi2w,4567
 verifiers/v1/gepa/__init__.py,sha256=6nmdRE0-34AKioPHBjTxUg5Jo_2z7tMX-OU3zMNpAJI,197
-verifiers/v1/gepa/adapter.py,sha256=yZBF_PBctEiBd0hgSt_jeL0biFY9kLl986mdtlet2s4,6199
+verifiers/v1/gepa/adapter.py,sha256=FivQBeUMvoCGaVqZUoLpj8S7GKPGyTuxKXZclnWHOKk,6185
 verifiers/v1/gepa/config.py,sha256=MVaJfPMehoPq5kF1vsm3lEJrxTiryesxHrQDcxE6TAE,4656
 verifiers/v1/gepa/dataset.py,sha256=N2QYqUljHVuAdCbUuydEIlarvAWa6LFdGKu9qsOvjro,2186
 verifiers/v1/gepa/reflection.py,sha256=T86Ar1BcSk012wfB1TioIP72wzBEEy8ZrwBy3rn2Hps,1045
@@ -319,7 +319,7 @@ verifiers/v1/utils/aio.py,sha256=ZKTHeURNbWhTpHMyYuqMLsdPWVHyn5anfG1IJs5y-Zg,148
 verifiers/v1/utils/compile.py,sha256=hSCCi06qIi4LH1wsAOgYVm0vma1tYZkCG-wLKq7PZxI,5724
 verifiers/v1/utils/format.py,sha256=NfQo9M5KMzWCZpQaet5YtsJx56vCKAZPfbjZZJLJhhM,2187
 verifiers/v1/utils/generic.py,sha256=xwu8xNGdVOhM2K87C5cqhmnRf6apSPDASA15M0S2MaY,2054
-verifiers/v1/utils/git.py,sha256=5VX5B3011yhclPex8Mc79ZmZ5ZT4hGfKaXoqL12QMdc,4872
+verifiers/v1/utils/git.py,sha256=NaVQ3_MG0x3nTb0MH2B-9O5xmnkPsaF5UGHRb4RIKvA,4837
 verifiers/v1/utils/image.py,sha256=OFw_wdwVtbdwa8uJ3X3vYfhpUAi0CTH5HMBupjVsRRQ,282
 verifiers/v1/utils/install.py,sha256=fWNsyKrw_PyhC0Qhqv5Ri0adFm5Ddx4AVy_hbsra1GU,1390
 verifiers/v1/utils/interrupt.py,sha256=F-KKhc5ndPJJfhd3SuMqyqhXhA32FhCRy5KWFJpEoM4,1179
@@ -327,8 +327,8 @@ verifiers/v1/utils/logging.py,sha256=OcMHA6NsYux3oIzjPuI95rDWmFBHNcHDjZeNIhXTX-Y
 verifiers/v1/utils/memory.py,sha256=ZkIvGk6uITAH5sKon65LifKPbvZr8mJ__PVc23FZpOQ,1835
 verifiers/v1/utils/sampling.py,sha256=JczGzBn6s3wsIrSY2Hy3hE8m6NreNgsNzhSjDikQqX0,1037
 verifiers/v1/utils/version.py,sha256=75ZtI2NHBmlb52KpcKLUpiSXp8q4bASr7uKeXoCKlT8,1582
-verifiers-0.2.2.dev31.dist-info/METADATA,sha256=Iny-bPcxK5EziGzvEeSvrnCFrIuDwh8rTDoaawejye0,4540
-verifiers-0.2.2.dev31.dist-info/WHEEL,sha256=lCkmxWfQsSc9CfIClYeavTdQeEX2toPqufh9gI35EQA,87
-verifiers-0.2.2.dev31.dist-info/entry_points.txt,sha256=dF82JUYEFslR1AOW8LZfuBWeZeQoyX4wCRrduklg8-I,551
-verifiers-0.2.2.dev31.dist-info/licenses/LICENSE,sha256=v0RrUsdV3IDoZhrRce297IXS3xMHNJ-_LdLpFAUWb9k,1072
-verifiers-0.2.2.dev31.dist-info/RECORD,,
+verifiers-0.2.2.dev33.dist-info/METADATA,sha256=IWBt9E4wd0eetYz62qsRhe4QkmKiJz0uyFZmSvQRMjI,4540
+verifiers-0.2.2.dev33.dist-info/WHEEL,sha256=lCkmxWfQsSc9CfIClYeavTdQeEX2toPqufh9gI35EQA,87
+verifiers-0.2.2.dev33.dist-info/entry_points.txt,sha256=dF82JUYEFslR1AOW8LZfuBWeZeQoyX4wCRrduklg8-I,551
+verifiers-0.2.2.dev33.dist-info/licenses/LICENSE,sha256=v0RrUsdV3IDoZhrRce297IXS3xMHNJ-_LdLpFAUWb9k,1072
+verifiers-0.2.2.dev33.dist-info/RECORD,,