Fix pytest

Author: rdheekonda

dreadnode/agent/agent.py

@@ -20,7 +20,6 @@
     tools_to_json_with_tag_transform,
 )
 
-from dreadnode import log_inputs, log_metric, log_outputs, score, task_span
 from dreadnode.agent.error import MaxStepsError
 from dreadnode.agent.events import (
     AgentEnd,
@@ -606,6 +605,8 @@ async def _process_tool_call(
         )
 
     def _log_event_metrics(self, event: AgentEvent) -> None:
+        from dreadnode import log_metric
+
         if isinstance(event, AgentEnd):
             log_metric("steps_taken", min(0, event.result.steps - 1))
             log_metric(f"stop_{event.stop_reason}", 1)
@@ -642,6 +643,8 @@ async def _stream_in_task(
         *,
         commit: CommitBehavior = "on-success",
     ) -> t.AsyncGenerator[AgentEvent, None]:
+        from dreadnode import log_inputs, log_outputs, score, task_span
+
         hooks = self._get_hooks()
         tool_names = [t.name for t in self.all_tools]
         stop_names = [s.name for s in self.stop_conditions]

dreadnode/agent/tools/tasking.py

@@ -1,6 +1,5 @@
 from loguru import logger
 
-from dreadnode import log_metric
 from dreadnode.agent.reactions import Fail, Finish
 from dreadnode.agent.tools.base import tool
 
@@ -28,6 +27,7 @@ async def finish_task(success: bool, summary: str) -> None:  # noqa: ARG001
     *   **Honest Status**: Accurately report the success or failure of the overall task. If any part of the task failed or was not completed, `success` should be `False`.
     *   **Comprehensive Summary**: The `summary` should be a complete and detailed markdown-formatted report of everything you did, including steps taken, tools used, and the final outcome. This is your final report to the user.
     """
+    from dreadnode import log_metric
 
     log_func = logger.success if success else logger.warning
     log_func(f"Agent finished the task (success={success})")
@@ -42,6 +42,7 @@ async def give_up_on_task(reason: str) -> None:  # noqa: ARG001
     """
     Give up on your task.
     """
+    from dreadnode import log_metric
 
     logger.info("Agent gave up on the task")
     log_metric("task_give_up", 1)

dreadnode/optimization/study.py

@@ -4,7 +4,6 @@
 
 from pydantic import ConfigDict, FilePath, PrivateAttr
 
-from dreadnode import log_inputs, log_metric, log_outputs, log_params, run, task_span
 from dreadnode.eval import Eval
 from dreadnode.eval.result import EvalResult
 from dreadnode.eval.sample import InputDataset
@@ -233,6 +232,8 @@ async def _stream(self) -> t.AsyncGenerator[StudyEvent[CandidateT], None]:  # no
         )
 
     def _log_event_metrics(self, event: StudyEvent[t.Any]) -> None:
+        from dreadnode import log_metric
+
         if isinstance(event, TrialComplete):
             trial = event.trial
             if trial.status == "success":
@@ -246,6 +247,8 @@ def _log_event_metrics(self, event: StudyEvent[t.Any]) -> None:
             log_metric("best_score", event.trial.score, step=event.trial.step)
 
     async def _stream_traced(self) -> t.AsyncGenerator[StudyEvent[CandidateT], None]:
+        from dreadnode import log_inputs, log_outputs, log_params, run, task_span
+
         objective_name = (
             self.objective
             if isinstance(self.objective, str)

dreadnode/scorers/classification.py

@@ -1,6 +1,6 @@
 import typing as t
 
-from transforms import pipeline  # type: ignore[import-not-found]
+from transformers import pipeline  # type: ignore[import-not-found]
 
 from dreadnode.meta import Config
 from dreadnode.metric import Metric

dreadnode/scorers/crucible.py

@@ -3,7 +3,6 @@
 
 import aiohttp
 
-from dreadnode import tag
 from dreadnode.metric import Metric
 from dreadnode.scorers import Scorer
 
@@ -31,6 +30,8 @@ def contains_crucible_flag(
     platform_url: str = "https://platform.dreadnode.io",
     name: str = "contains_crucible_flag",
 ) -> Scorer[t.Any]:
+    from dreadnode import tag
+
     async def evaluate(
         obj: t.Any,
         *,

dreadnode/task.py

@@ -6,9 +6,6 @@
 import typing_extensions as te
 from opentelemetry.trace import Tracer
 
-from dreadnode import score
-from dreadnode.airt.target import CustomTarget
-from dreadnode.eval.eval import Eval
 from dreadnode.meta.context import Context
 from dreadnode.meta.types import Component, ConfigInfo
 from dreadnode.scorers.base import Scorer, ScorerCallable, ScorersLike
@@ -23,7 +20,9 @@
 )
 
 if t.TYPE_CHECKING:
+    from dreadnode.airt.target.custom import CustomTarget
     from dreadnode.eval.eval import (
+        Eval,
         InputDataset,
         InputDatasetProcessor,
     )
@@ -379,6 +378,8 @@ def as_eval(
         scorers: "ScorersLike[R] | None" = None,
         assert_scores: list[str] | t.Literal[True] | None = None,
     ) -> "Eval[t.Any, R]":
+        from dreadnode.eval.eval import Eval
+
         if isinstance(dataset, str):
             dataset = Path(dataset)
 
@@ -402,6 +403,8 @@ def as_target(
         self,
         input_param_name: str | None = None,
     ) -> "CustomTarget[R]":
+        from dreadnode.airt.target.custom import CustomTarget
+
         return CustomTarget(task=self, input_param_name=input_param_name)
 
     async def run_always(self, *args: P.args, **kwargs: P.kwargs) -> TaskSpan[R]:
@@ -417,6 +420,7 @@ async def run_always(self, *args: P.args, **kwargs: P.kwargs) -> TaskSpan[R]:
         Returns:
             The span associated with task execution.
         """
+        from dreadnode import score
 
         run = current_run_span.get()