verifiers 0.1.15.dev423__tar.gz → 0.1.15.dev424__tar.gz

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: verifiers
-Version: 0.1.15.dev423
+Version: 0.1.15.dev424
 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-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/__init__.py

@@ -12,7 +12,7 @@ from verifiers.v1.clients import (
     BaseClientConfig,
     Client,
     ClientConfig,
-    RolloutContext,
+    ModelContext,
     resolve_client,
 )
 from verifiers.v1.decorators import group_reward, metric, reward, stop, tool
@@ -188,7 +188,7 @@ __all__ = [
     "BaseConfig",
     "Harness",
     "HarnessConfig",
-    "RolloutContext",
+    "ModelContext",
     "Runtime",
     "RuntimeConfig",
     "ProgramResult",

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/cli/eval/runner.py

@@ -6,7 +6,7 @@ import logging
 import random
 import time
 
-from verifiers.v1.clients import RolloutContext, resolve_client
+from verifiers.v1.clients import ModelContext, resolve_client
 from verifiers.v1.configs.eval import EvalConfig
 from verifiers.v1.cli.eval import resume
 from verifiers.v1.cli.dashboard import dashboard
@@ -29,7 +29,7 @@ async def run_eval(env: Environment, config: EvalConfig) -> list[Trace]:
     if config.shuffle:
         random.Random(_SHUFFLE_SEED).shuffle(tasks)
     tasks = tasks if config.num_tasks is None else tasks[: config.num_tasks]
-    ctx = RolloutContext(client=client, model=config.model, sampling=config.sampling)
+    ctx = ModelContext(client=client, model=config.model, sampling=config.sampling)
     # One episode of `num_rollouts` rollouts per task; the shared semaphore bounds total
     # concurrent rollouts (across episodes), so group rewards still see their whole episode.
     semaphore = (

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/cli/init.py

@@ -192,7 +192,7 @@ class {prefix}Harness(vf.Harness[{prefix}HarnessConfig]):
 
     async def launch(
         self,
-        ctx: vf.RolloutContext,
+        ctx: vf.ModelContext,
         trace: vf.Trace,
         runtime: vf.Runtime,
         endpoint: str,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/clients/__init__.py

@@ -1,6 +1,6 @@
 """The client abstraction and its OpenAI-compatible + renderer implementations."""
 
-from verifiers.v1.clients.client import Client, RolloutContext
+from verifiers.v1.clients.client import Client, ModelContext
 from verifiers.v1.clients.config import (
     BaseClientConfig,
     ClientConfig,
@@ -13,7 +13,7 @@ from verifiers.v1.clients.train import TrainClient
 
 __all__ = [
     "Client",
-    "RolloutContext",
+    "ModelContext",
     "BaseClientConfig",
     "ClientConfig",
     "EvalClientConfig",

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/clients/client.py

@@ -77,9 +77,9 @@ class Client(ABC):
 
 
 @dataclass(frozen=True)
-class RolloutContext:
-    """The collaborators a single rollout needs (client + model + sampling), bundled
-    so harnesses hold no rollout state. Built by the Environment."""
+class ModelContext:
+    """The model-side collaborators a harness talks to (client + model + sampling),
+    bundled so harnesses hold no rollout state. Built by the Environment."""
 
     model: str
     client: Client

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/env.py

@@ -19,7 +19,7 @@ from pydantic import Field, SerializeAsAny, model_validator
 from pydantic_config import BaseConfig
 
 from verifiers.v1.harness import HarnessConfig
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.decorators import discover_decorated
 from verifiers.v1.episode import Episode
 from verifiers.v1.types import ID
@@ -300,7 +300,7 @@ class Environment:
             self.harness.config.runtime, task, self._warned_resources
         )
 
-    def episode(self, task: Task, ctx: RolloutContext, n: int = 1) -> Episode:
+    def episode(self, task: Task, ctx: ModelContext, n: int = 1) -> Episode:
         """Resolve `task` into a runnable episode of `n` rollouts: pick its runtime
         (image + resources) and its timeouts (cli/toml > task > default, None = no limit),
         build one `Rollout` per sample sharing them, and wrap them in an `Episode` (which

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/harness.py

@@ -16,7 +16,7 @@ from typing import ClassVar, Generic, TypeVar
 from pydantic import Field
 from pydantic_config import BaseConfig
 
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.decorators import discover_decorated, invoke
 from verifiers.v1.errors import HarnessError, boundary
 from verifiers.v1.utils.install import env_name
@@ -131,7 +131,7 @@ class Harness(ABC, Generic[ConfigT]):
 
     async def run(
         self,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         trace: Trace,
         runtime: Runtime,
         endpoint: str,
@@ -176,7 +176,7 @@ class Harness(ABC, Generic[ConfigT]):
     @abstractmethod
     async def launch(
         self,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         trace: Trace,
         runtime: Runtime,
         endpoint: str,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/harnesses/codex/harness.py

@@ -10,7 +10,7 @@ session secret) is read from an env var.
 import logging
 import shlex
 
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.harness import Harness, HarnessConfig
 from verifiers.v1.runtimes import ProgramResult, Runtime
 from verifiers.v1.trace import Trace
@@ -65,7 +65,7 @@ class CodexHarness(Harness[CodexHarnessConfig]):
 
     async def launch(
         self,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         trace: Trace,
         runtime: Runtime,
         endpoint: str,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/harnesses/default/harness.py

@@ -14,7 +14,7 @@ import os
 from pathlib import Path
 
 from verifiers.v1.harness import Harness, HarnessConfig
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.dialects.chat import message_to_wire
 from verifiers.v1.runtimes import ProgramResult, Runtime
 from verifiers.v1.trace import Trace
@@ -61,7 +61,7 @@ class DefaultHarness(Harness[DefaultHarnessConfig]):
 
     async def launch(
         self,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         trace: Trace,
         runtime: Runtime,
         endpoint: str,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/harnesses/kimi_code/harness.py

@@ -9,7 +9,7 @@ import json
 import logging
 import shlex
 
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.harness import Harness, HarnessConfig
 from verifiers.v1.runtimes import ProgramResult, Runtime
 from verifiers.v1.trace import Trace
@@ -64,7 +64,7 @@ class KimiCodeHarness(Harness[KimiCodeHarnessConfig]):
 
     async def launch(
         self,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         trace: Trace,
         runtime: Runtime,
         endpoint: str,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/harnesses/mini_swe_agent/harness.py

@@ -2,7 +2,7 @@
 
 from pathlib import Path
 
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.harness import Harness, HarnessConfig
 from verifiers.v1.runtimes import ProgramResult, Runtime
 from verifiers.v1.trace import Trace
@@ -27,7 +27,7 @@ class MiniSWEAgentHarness(Harness[MiniSWEAgentHarnessConfig]):
 
     async def launch(
         self,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         trace: Trace,
         runtime: Runtime,
         endpoint: str,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/harnesses/null/harness.py

@@ -11,7 +11,7 @@ import json
 from pathlib import Path
 
 from verifiers.v1.harness import Harness, HarnessConfig
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.dialects.chat import message_to_wire
 from verifiers.v1.runtimes import ProgramResult, Runtime
 from verifiers.v1.trace import Trace
@@ -35,7 +35,7 @@ class NullHarness(Harness[NullHarnessConfig]):
 
     async def launch(
         self,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         trace: Trace,
         runtime: Runtime,
         endpoint: str,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/harnesses/rlm/harness.py

@@ -18,7 +18,7 @@ from typing import Literal
 from pydantic import model_validator
 
 from verifiers.v1.harness import Harness, HarnessConfig
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.decorators import metric
 from verifiers.v1.runtimes import ProgramResult, Runtime
 from verifiers.v1.trace import Trace
@@ -115,7 +115,7 @@ class RLMHarness(Harness[RLMHarnessConfig]):
 
     async def launch(
         self,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         trace: Trace,
         runtime: Runtime,
         endpoint: str,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/harnesses/terminus_2/harness.py

@@ -3,7 +3,7 @@
 import logging
 from pathlib import Path
 
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.harness import Harness, HarnessConfig
 from verifiers.v1.runtimes import ProgramResult, Runtime
 from verifiers.v1.trace import Trace
@@ -39,7 +39,7 @@ class Terminus2Harness(Harness[Terminus2HarnessConfig]):
 
     async def launch(
         self,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         trace: Trace,
         runtime: Runtime,
         endpoint: str,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/interception/server.py

@@ -32,7 +32,7 @@ from aiohttp import web
 from pydantic import TypeAdapter, ValidationError
 from pydantic_core import PydanticSerializationError, from_json, to_json
 
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.dialects import DIALECTS, Dialect
 from verifiers.v1.dialects.base import is_sse_done_event
 from verifiers.v1 import graph
@@ -129,7 +129,7 @@ class RolloutSession:
     checked before each turn, and (optionally) a user simulator the rollout sets before the
     harness runs."""
 
-    ctx: RolloutContext
+    ctx: ModelContext
     trace: Trace
     stops: list[Callable[[Trace], Awaitable[bool]]] = field(default_factory=list)
     limits: RolloutLimits = field(default_factory=RolloutLimits)

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/rollout.py

@@ -19,7 +19,7 @@ from contextlib import asynccontextmanager
 from enum import StrEnum
 
 from verifiers.v1.harness import Harness
-from verifiers.v1.clients import RolloutContext
+from verifiers.v1.clients import ModelContext
 from verifiers.v1.decorators import discover_decorated, invoke
 from verifiers.v1.errors import (
     HarnessError,
@@ -69,7 +69,7 @@ class Rollout:
         task: Task,
         taskset: Taskset,
         harness: Harness,
-        ctx: RolloutContext,
+        ctx: ModelContext,
         runtime_config: RuntimeConfig,
         setup_timeout: float | None = None,
         harness_timeout: float | None = None,

{verifiers-0.1.15.dev423 → verifiers-0.1.15.dev424}/verifiers/v1/serve/server.py

@@ -6,7 +6,7 @@ loads it. A caller (e.g. the orchestrator) stays env-agnostic: it asks `info` fo
 the task count + whether group scoring is needed, then `run_rollout` / `run_group`
 by task index. Per request the server resolves a `Client` from the request's
 `client` config (cached, so a renderer's tokenizer is built once), wraps it in a
-`RolloutContext`, and runs `env.episode(task, ctx, n).run()`, returning each
+`ModelContext`, and runs `env.episode(task, ctx, n).run()`, returning each
 `Trace` as a JSON dict (with its computed `branches`).
 
 Minimal port of `verifiers.serve` (ROUTER + msgpack), single async process: each
@@ -26,7 +26,7 @@ import zmq.asyncio
 
 from verifiers.utils.process_utils import use_threading_tqdm_lock
 from verifiers.utils.serve_utils import msgpack_encoder
-from verifiers.v1.clients import RolloutContext, resolve_client
+from verifiers.v1.clients import ModelContext, resolve_client
 from verifiers.v1.clients.client import Client
 from verifiers.v1.clients.config import ClientConfig
 from verifiers.v1.decorators import discover_decorated
@@ -106,8 +106,8 @@ class EnvServer:
 
     def _context(
         self, client_config: ClientConfig, model: str, sampling: SamplingConfig
-    ) -> RolloutContext:
-        return RolloutContext(
+    ) -> ModelContext:
+        return ModelContext(
             client=self._client(client_config, model), model=model, sampling=sampling
         )