minor fixes

README.md

@@ -2,7 +2,7 @@
 title: SQLSherlock Env
 emoji: 🔍
 colorFrom: indigo
-colorTo: cyan
+colorTo: blue
 sdk: docker
 app_port: 7860
 tags:

inference.py

@@ -45,11 +45,11 @@ SPACE_URL    = os.getenv("SPACE_URL",    "http://localhost:7860")
 # Optional — if you use from_docker_image():
 LOCAL_IMAGE_NAME = os.getenv("LOCAL_IMAGE_NAME")
 
-# Use full environment max_steps — no artificial restriction
+# Full environment max_steps — agent gets maximum room to clean
 STEP_BUDGETS: dict[str, int] = {
-    "task1_null_and_types":         20,   # env max_steps = 20
-    "task2_constraints_and_fk":     25,   # env max_steps = 25
-    "task3_full_audit_with_trap":   30,   # env max_steps = 30
+    "task1_null_and_types":         30,   # env max_steps = 30
+    "task2_constraints_and_fk":     40,   # env max_steps = 40
+    "task3_full_audit_with_trap":   50,   # env max_steps = 50
 }
 
 TASKS = [

openenv.yaml

@@ -24,7 +24,7 @@ tasks:
   - id: task1_null_and_types
     name: "Null and type error repair"
     difficulty: easy
-    max_steps: 20
+    max_steps: 30
     description: >
       Find and fix null values and type errors in the primary table.
       Profile columns, identify anomalies, fix with reasoning,
@@ -33,7 +33,7 @@ tasks:
   - id: task2_constraints_and_fk
     name: "Constraint and FK integrity"
     difficulty: medium
-    max_steps: 25
+    max_steps: 40
     description: >
       Everything in Task 1 plus constraint violations
       (negative values in must-be-positive columns) and FK
@@ -42,7 +42,7 @@ tasks:
   - id: task3_full_audit_with_trap
     name: "Full statistical audit with trap"
     difficulty: hard
-    max_steps: 30
+    max_steps: 50
     description: >
       Full audit including statistical outliers. TRAP WARNING:
       one numeric value looks suspicious but is legitimate.

sqlsherlock_env/server/environment.py

@@ -33,7 +33,7 @@ TASKS: list[dict] = [
         "id":          "task1_null_and_types",
         "name":        "Null and type error repair",
         "difficulty":  "easy",
-        "max_steps":   20,
+        "max_steps":   30,
         "description": (
             "Find and fix null values and type errors in the primary table. "
             "Profile columns, identify anomalies, fix with reasoning, "
@@ -44,7 +44,7 @@ TASKS: list[dict] = [
         "id":          "task2_constraints_and_fk",
         "name":        "Constraint and FK integrity",
         "difficulty":  "medium",
-        "max_steps":   25,
+        "max_steps":   40,
         "description": (
             "Everything in Task 1 plus constraint violations "
             "(negative values in must-be-positive columns) and FK "
@@ -55,7 +55,7 @@ TASKS: list[dict] = [
         "id":          "task3_full_audit_with_trap",
         "name":        "Full statistical audit with trap",
         "difficulty":  "hard",
-        "max_steps":   30,
+        "max_steps":   50,
         "description": (
             "Full audit including statistical outliers. TRAP WARNING: "
             "one numeric value looks suspicious but is legitimate. "
@@ -100,21 +100,10 @@ class SQLSherlockEnvironment(Environment):
         Raises:
             ValueError: If dataset or task_id is missing/invalid.
         """
-        dataset = kwargs.get("dataset", "")
-        task_id = kwargs.get("task_id", "")
-        seed    = int(kwargs.get("seed", 42))
+        dataset  = kwargs.get("dataset", "") or "phihung/titanic"
+        task_id  = kwargs.get("task_id", "") or "task1_null_and_types"
+        seed     = int(kwargs.get("seed", 42))
         max_rows = int(kwargs.get("max_rows", 500))
-
-        if not dataset or not dataset.strip():
-            raise ValueError(
-                "reset() requires 'dataset' keyword argument. "
-                "Provide a file path, HuggingFace dataset name, or raw CSV text."
-            )
-        if not task_id or not task_id.strip():
-            raise ValueError(
-                "reset() requires 'task_id' keyword argument. "
-                f"Valid tasks: {sorted(_TASK_MAP.keys())}"
-            )
         if task_id not in _TASK_MAP:
             raise ValueError(
                 f"Unknown task_id '{task_id}'. "

tests/test_environment.py

@@ -73,9 +73,9 @@ class TestTasksCatalogue:
 
     def test_max_steps_values(self):
         step_map = {t["id"]: t["max_steps"] for t in TASKS}
-        assert step_map["task1_null_and_types"]       == 20
-        assert step_map["task2_constraints_and_fk"]   == 25
-        assert step_map["task3_full_audit_with_trap"] == 30
+        assert step_map["task1_null_and_types"]       == 30
+        assert step_map["task2_constraints_and_fk"]   == 40
+        assert step_map["task3_full_audit_with_trap"] == 50
 
 
 # ---------------------------------------------------------------------------
@@ -99,13 +99,16 @@ class TestReset:
         obs = env.reset(dataset=RAW_CSV_TEXT, task_id="task1_null_and_types")
         assert obs.step == 0
 
-    def test_reset_no_dataset_raises(self, env):
-        with pytest.raises(ValueError, match="dataset"):
-            env.reset(dataset="", task_id="task1_null_and_types")
+    def test_reset_no_dataset_uses_default(self, env):
+        """Empty dataset defaults to phihung/titanic."""
+        obs = env.reset(dataset="", task_id="task1_null_and_types")
+        assert isinstance(obs, SQLSherlockObservation)
+        assert len(obs.tables_summary) > 0
 
-    def test_reset_no_task_raises(self, env):
-        with pytest.raises(ValueError, match="task_id"):
-            env.reset(dataset=RAW_CSV_TEXT, task_id="")
+    def test_reset_no_task_uses_default(self, env):
+        """Empty task_id defaults to task1_null_and_types."""
+        obs = env.reset(dataset=RAW_CSV_TEXT, task_id="")
+        assert isinstance(obs, SQLSherlockObservation)
 
     def test_reset_invalid_task_raises(self, env):
         with pytest.raises(ValueError, match="Unknown task_id"):
@@ -391,7 +394,7 @@ class TestMaxSteps:
         env.reset(dataset=RAW_CSV_TEXT, task_id="task1_null_and_types")
         table = list(env._db.table_names())[0]
         done = False
-        for _ in range(25):   # more than max_steps=20
+        for _ in range(35):   # more than max_steps=30
             _, _, done, _ = _step(env,
                 SQLSherlockAction(action_type="inspect", table=table)
             )