Fix/misalignment qa housekeeping

aieng-eval-agents/aieng/agent_evals/misalignment_qa/experiment.py

@@ -25,7 +25,6 @@
     prepare_variant_runs,
 )
 from aieng.agent_evals.misalignment_qa.task import MisalignmentTask
-from dotenv import load_dotenv
 
 
 logger = logging.getLogger(__name__)
@@ -313,8 +312,6 @@ def select_variant_runs(
 
 async def run_experiment_config(config: ExperimentConfig, *, variant_ids: set[str] | None = None) -> None:
     """Run the full experiment: upload dataset, iterate variants, collect warnings."""
-    load_dotenv(verbose=True)
-
     prepared_tasks = prepare_dataset_items(config)
     execution = create_execution_identity()
     prepared_variants = select_variant_runs(prepare_variant_runs(config, execution=execution), variant_ids=variant_ids)

aieng-eval-agents/aieng/agent_evals/misalignment_qa/task.py

@@ -6,6 +6,7 @@
 from collections.abc import Sequence
 from typing import Any
 
+from google.adk.agents import LlmAgent
 from google.adk.events import Event
 from google.adk.runners import Runner
 from google.adk.sessions import InMemorySessionService
@@ -35,7 +36,7 @@ class MisalignmentTask:
     def __init__(
         self,
         *,
-        agent: Any,
+        agent: LlmAgent,
         shared_turns: Sequence[dict[str, Any]] | None = None,
         user_context_preamble: str | None = None,
     ) -> None:

implementations/misalignment_qa/evaluate.py

@@ -9,6 +9,7 @@
 from pathlib import Path
 
 from aieng.agent_evals.misalignment_qa.experiment import load_experiment_config, run_experiment_config
+from dotenv import load_dotenv
 
 
 def main() -> None:
@@ -24,6 +25,8 @@ def main() -> None:
     parser.add_argument("--log-level", default="INFO", type=str, help="Logging level (e.g. INFO, DEBUG).")
     args = parser.parse_args()
 
+    load_dotenv(verbose=True)
+
     logging.basicConfig(
         level=getattr(logging, args.log_level.upper(), logging.INFO),
         format="%(asctime)s %(levelname)s %(name)s: %(message)s",