fix: implement abstract methods in SQLAnalyticsEnv and use sync() wrapper

- Implement _step_payload, _parse_result, _parse_state in client.py
- Use .sync() wrapper since EnvClient is async-first
- Access StepResult.observation for env-specific fields (db_schema, error, result)
- Access StepResult.reward and StepResult.done for step-level fields

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

client.py

@@ -36,7 +36,9 @@ Example usage
         env.close()
 """
 
-from openenv.core.env_client import EnvClient
+from typing import Any, Dict
+
+from openenv.core.env_client import EnvClient, StepResult
 
 try:
     from .models import SQLAction, SQLObservation, SQLState
@@ -49,8 +51,8 @@ class SQLAnalyticsEnv(EnvClient[SQLAction, SQLObservation, SQLState]):
     Typed client for the SQL Analytics OpenEnv environment.
 
     Inherits the full EnvClient interface:
-        reset(**kwargs) -> SQLObservation
-        step(action: SQLAction) -> SQLObservation
+        reset(**kwargs) -> StepResult[SQLObservation]
+        step(action: SQLAction) -> StepResult[SQLObservation]
         state() -> SQLState
         close()
 
@@ -58,4 +60,18 @@ class SQLAnalyticsEnv(EnvClient[SQLAction, SQLObservation, SQLState]):
         SQLAnalyticsEnv.from_env(hf_repo_id)
         SQLAnalyticsEnv.from_docker_image(image_tag)
     """
-    pass
+
+    def _step_payload(self, action: SQLAction) -> Dict[str, Any]:
+        return action.model_dump()
+
+    def _parse_result(self, payload: Dict[str, Any]) -> StepResult[SQLObservation]:
+        obs_dict = payload.get("observation", {})
+        obs = SQLObservation(**obs_dict)
+        return StepResult(
+            observation=obs,
+            reward=payload.get("reward"),
+            done=payload.get("done", False),
+        )
+
+    def _parse_state(self, payload: Dict[str, Any]) -> SQLState:
+        return SQLState(**payload)

inference.py

@@ -141,11 +141,11 @@ def _extract_sql(text: str) -> str:
 
 # ── Agent loop ────────────────────────────────────────────────────────────────
 
-def run_task(client: OpenAI, env: SQLAnalyticsEnv, task_id: str) -> float:
+def run_task(client: OpenAI, env, task_id: str) -> float:
     """Run one episode. Returns the final score in [0, 1]."""
-    obs = env.reset(task_id=task_id)
-    schema           = obs.db_schema or ""
-    task_description = obs.task_description or ""
+    result = env.reset(task_id=task_id)
+    schema           = result.observation.db_schema or ""
+    task_description = result.observation.task_description or ""
 
     log_start(task=task_id, env=BENCHMARK, model=MODEL_NAME)
 
@@ -186,12 +186,12 @@ def run_task(client: OpenAI, env: SQLAnalyticsEnv, task_id: str) -> float:
                 print(f"[DEBUG] No SQL at step={step}, stopping.", file=sys.stderr, flush=True)
                 break
 
-            obs = env.step(SQLAction(sql=sql))
+            result = env.step(SQLAction(sql=sql))
 
-            reward      = obs.reward or 0.0
-            done        = obs.done
-            error       = obs.error
-            result_rows = len(obs.result) if obs.result else 0
+            reward      = result.reward or 0.0
+            done        = result.done
+            error       = result.observation.error
+            result_rows = len(result.observation.result) if result.observation.result else 0
             steps_taken = step
 
             rewards.append(reward)
@@ -223,7 +223,7 @@ def run_task(client: OpenAI, env: SQLAnalyticsEnv, task_id: str) -> float:
 
 def main() -> None:
     client = OpenAI(base_url=API_BASE_URL, api_key=API_KEY)
-    env    = SQLAnalyticsEnv(base_url=ENV_BASE_URL)
+    env    = SQLAnalyticsEnv(base_url=ENV_BASE_URL).sync()
 
     scores = {}
     try: