diff --git a/tests/test_dict_flattening.py b/tests/test_dict_flattening.py
index cd4776d..9bad7bf 100644
--- a/tests/test_dict_flattening.py
+++ b/tests/test_dict_flattening.py
@@ -18,9 +18,6 @@
 from fz.io import flatten_dict_recursive, flatten_dict_columns
 
 
-# Skip all tests if pandas is not available (dict flattening requires pandas)
-
-
 class TestFlattenDictRecursive:
     """Test the flatten_dict_recursive helper function"""
 
@@ -211,69 +208,81 @@ class TestFzoWithDictFlattening:
     def test_fzo_with_dict_output(self):
         """Test fzo automatically flattens dict outputs"""
         with tempfile.TemporaryDirectory() as tmpdir:
-            # Create result directory with dict output
-            result_dir = Path(tmpdir) / "results" / "x=5,y=10"
-            result_dir.mkdir(parents=True)
-
-            # Write output file with JSON dict
-            with open(result_dir / "output.txt", "w") as f:
-                f.write("sum=15\n")
-                f.write('stats={"min": 5, "max": 10, "diff": 5}\n')
-
-            # Define model
-            model = {
-                "varprefix": "$",
-                "delim": "{}",
-                "output": {
-                    "sum": "grep 'sum=' output.txt | cut -d'=' -f2",
-                    "stats": "grep 'stats=' output.txt | cut -d'=' -f2"
+            # Save original directory to avoid Windows file deletion issues
+            original_cwd = os.getcwd()
+            try:
+                # Create result directory with dict output
+                result_dir = Path(tmpdir) / "results" / "x=5,y=10"
+                result_dir.mkdir(parents=True)
+
+                # Write output file with JSON dict
+                with open(result_dir / "output.txt", "w") as f:
+                    f.write("sum=15\n")
+                    f.write('stats={"min": 5, "max": 10, "diff": 5}\n')
+
+                # Define model
+                model = {
+                    "varprefix": "$",
+                    "delim": "{}",
+                    "output": {
+                        "sum": "grep 'sum=' output.txt | cut -d'=' -f2",
+                        "stats": "grep 'stats=' output.txt | cut -d'=' -f2"
+                    }
                 }
-            }
-
-            # Run fzo
-            os.chdir(tmpdir)
-            results = fz.fzo("results/*", model)
-
-            # Check flattening occurred
-            assert 'stats' not in results.columns
-            assert 'stats_min' in results.columns
-            assert 'stats_max' in results.columns
-            assert 'stats_diff' in results.columns
 
-            # Check values
-            assert results['sum'].iloc[0] == 15
-            assert results['stats_min'].iloc[0] == 5
-            assert results['stats_max'].iloc[0] == 10
-            assert results['stats_diff'].iloc[0] == 5
+                # Run fzo
+                os.chdir(tmpdir)
+                results = fz.fzo("results/*", model)
+
+                # Check flattening occurred
+                assert 'stats' not in results.columns
+                assert 'stats_min' in results.columns
+                assert 'stats_max' in results.columns
+                assert 'stats_diff' in results.columns
+
+                # Check values
+                assert results['sum'].iloc[0] == 15
+                assert results['stats_min'].iloc[0] == 5
+                assert results['stats_max'].iloc[0] == 10
+                assert results['stats_diff'].iloc[0] == 5
+            finally:
+                # Restore original directory to allow cleanup on Windows
+                os.chdir(original_cwd)
 
     def test_fzo_with_nested_dict_output(self):
         """Test fzo with nested dict outputs"""
         with tempfile.TemporaryDirectory() as tmpdir:
-            result_dir = Path(tmpdir) / "results" / "case1"
-            result_dir.mkdir(parents=True)
-
-            # Write output with nested dict
-            nested_dict = {
-                'basic': {'min': 1, 'max': 10},
-                'advanced': {'mean': 5.5, 'std': 2.5}
-            }
-            with open(result_dir / "output.txt", "w") as f:
-                f.write(f"data={json.dumps(nested_dict)}\n")
+            # Save original directory to avoid Windows file deletion issues
+            original_cwd = os.getcwd()
+            try:
+                result_dir = Path(tmpdir) / "results" / "case1"
+                result_dir.mkdir(parents=True)
+
+                # Write output with nested dict
+                nested_dict = {
+                    'basic': {'min': 1, 'max': 10},
+                    'advanced': {'mean': 5.5, 'std': 2.5}
+                }
+                with open(result_dir / "output.txt", "w") as f:
+                    f.write(f"data={json.dumps(nested_dict)}\n")
 
-            model = {
-                "output": {
-                    "data": "grep 'data=' output.txt | cut -d'=' -f2"
+                model = {
+                    "output": {
+                        "data": "grep 'data=' output.txt | cut -d'=' -f2"
+                    }
                 }
-            }
 
-            os.chdir(tmpdir)
-            results = fz.fzo("results/*", model)
+                os.chdir(tmpdir)
+                results = fz.fzo("results/*", model)
 
-            # Check nested flattening
-            assert 'data_basic_min' in results.columns
-            assert 'data_basic_max' in results.columns
-            assert 'data_advanced_mean' in results.columns
-            assert 'data_advanced_std' in results.columns
+                # Check nested flattening
+                assert 'data_basic_min' in results.columns
+                assert 'data_basic_max' in results.columns
+                assert 'data_advanced_mean' in results.columns
+                assert 'data_advanced_std' in results.columns
+            finally:
+                # Restore original directory to allow cleanup on Windows
+                os.chdir(original_cwd)
 
 
 class TestFzrWithDictFlattening:
@@ -282,16 +291,19 @@ class TestFzrWithDictFlattening:
     def test_fzr_with_dict_output(self):
         """Test fzr automatically flattens dict outputs"""
         with tempfile.TemporaryDirectory() as tmpdir:
-            os.chdir(tmpdir)
-
-            # Create input template
-            with open("input.txt", "w") as f:
-                f.write("x = ${x}\n")
-
-            # Create calculator script that produces dict output
-            calc_script = Path(tmpdir) / "calc.py"
-            with open(calc_script, "w") as f:
-                f.write("""#!/usr/bin/env python3
+            # Save original directory to avoid Windows file deletion issues
+            original_cwd = os.getcwd()
+            try:
+                os.chdir(tmpdir)
+
+                # Create input template
+                with open("input.txt", "w") as f:
+                    f.write("x = ${x}\n")
+
+                # Create calculator script that produces dict output
+                calc_script = Path(tmpdir) / "calc.py"
+                with open(calc_script, "w") as f:
+                    f.write("""#!/usr/bin/env python3
 import json
 
 # Read input
@@ -307,53 +319,59 @@ def test_fzr_with_dict_output(self):
     f.write(f"value={x}\\n")
     f.write(f"stats={json.dumps(stats)}\\n")
 """)
-            os.chmod(calc_script, 0o755)
-
-            # Define model
-            model = {
-                "varprefix": "$",
-                "delim": "{}",
-                "output": {
-                    "value": "grep 'value=' output.txt | cut -d'=' -f2",
-                    "stats": "grep 'stats=' output.txt | cut -d'=' -f2"
+                os.chmod(calc_script, 0o755)
+
+                # Define model
+                model = {
+                    "varprefix": "$",
+                    "delim": "{}",
+                    "output": {
+                        "value": "grep 'value=' output.txt | cut -d'=' -f2",
+                        "stats": "grep 'stats=' output.txt | cut -d'=' -f2"
+                    }
                 }
-            }
 
-            # Run fzr
-            results = fz.fzr(
-                input_path="input.txt",
-                input_variables={"x": [5, 10, 15]},
-                model=model,
-                calculators=f"sh://python3 {calc_script}"
-            )
-
-            # Check flattening occurred
-            assert 'stats' not in results.columns
-            assert 'stats_min' in results.columns
-            assert 'stats_max' in results.columns
-            assert 'stats_mean' in results.columns
-
-            # Check values for first row
-            assert results['x'].iloc[0] == 5
-            assert results['value'].iloc[0] == 5
-            assert results['stats_min'].iloc[0] == 4
-            assert results['stats_max'].iloc[0] == 6
-            assert results['stats_mean'].iloc[0] == 5
-
-            # Check all rows
-            assert len(results) == 3
+                # Run fzr
+                results = fz.fzr(
+                    input_path="input.txt",
+                    input_variables={"x": [5, 10, 15]},
+                    model=model,
+                    calculators=f"sh://python3 {calc_script}"
+                )
+
+                # Check flattening occurred
+                assert 'stats' not in results.columns
+                assert 'stats_min' in results.columns
+                assert 'stats_max' in results.columns
+                assert 'stats_mean' in results.columns
+
+                # Check values for first row
+                assert results['x'].iloc[0] == 5
+                assert results['value'].iloc[0] == 5
+                assert results['stats_min'].iloc[0] == 4
+                assert results['stats_max'].iloc[0] == 6
+                assert results['stats_mean'].iloc[0] == 5
+
+                # Check all rows
+                assert len(results) == 3
+            finally:
+                # Restore original directory to allow cleanup on Windows
+                os.chdir(original_cwd)
 
     def test_fzr_with_deeply_nested_dict(self):
         """Test fzr with deeply nested dict outputs (3 levels)"""
         with tempfile.TemporaryDirectory() as tmpdir:
-            os.chdir(tmpdir)
+            # Save original directory to avoid Windows file deletion issues
+            original_cwd = os.getcwd()
+            try:
+                os.chdir(tmpdir)
 
-            with open("input.txt", "w") as f:
-                f.write("x = ${x}\n")
+                with open("input.txt", "w") as f:
+                    f.write("x = ${x}\n")
 
-            calc_script = Path(tmpdir) / "calc.py"
-            with open(calc_script, "w") as f:
-                f.write("""#!/usr/bin/env python3
+                calc_script = Path(tmpdir) / "calc.py"
+                with open(calc_script, "w") as f:
+                    f.write("""#!/usr/bin/env python3
 import json
 
 with open('input.txt', 'r') as f:
@@ -375,44 +393,50 @@ def test_fzr_with_deeply_nested_dict(self):
 with open('output.txt', 'w') as f:
     f.write(f"result={json.dumps(result)}\\n")
 """)
-            os.chmod(calc_script, 0o755)
-
-            model = {
-                "varprefix": "$",
-                "delim": "{}",
-                "output": {
-                    "result": "grep 'result=' output.txt | cut -d'=' -f2"
+                os.chmod(calc_script, 0o755)
+
+                model = {
+                    "varprefix": "$",
+                    "delim": "{}",
+                    "output": {
+                        "result": "grep 'result=' output.txt | cut -d'=' -f2"
+                    }
                 }
-            }
-
-            results = fz.fzr(
-                input_path="input.txt",
-                input_variables={"x": [3, 5]},
-                model=model,
-                calculators=f"sh://python3 {calc_script}"
-            )
 
-            # Check deep nesting flattened correctly
-            assert 'result_level1_level2_level3_value' in results.columns
-            assert 'result_level1_level2_level3_squared' in results.columns
-
-            # Check values
-            assert results['result_level1_level2_level3_value'].iloc[0] == 6
-            assert results['result_level1_level2_level3_squared'].iloc[0] == 9
-            assert results['result_level1_level2_level3_value'].iloc[1] == 10
-            assert results['result_level1_level2_level3_squared'].iloc[1] == 25
+                results = fz.fzr(
+                    input_path="input.txt",
+                    input_variables={"x": [3, 5]},
+                    model=model,
+                    calculators=f"sh://python3 {calc_script}"
+                )
+
+                # Check deep nesting flattened correctly
+                assert 'result_level1_level2_level3_value' in results.columns
+                assert 'result_level1_level2_level3_squared' in results.columns
+
+                # Check values
+                assert results['result_level1_level2_level3_value'].iloc[0] == 6
+                assert results['result_level1_level2_level3_squared'].iloc[0] == 9
+                assert results['result_level1_level2_level3_value'].iloc[1] == 10
+                assert results['result_level1_level2_level3_squared'].iloc[1] == 25
+            finally:
+                # Restore original directory to allow cleanup on Windows
+                os.chdir(original_cwd)
 
     def test_fzr_with_multiple_dict_outputs(self):
         """Test fzr with multiple dict-valued outputs"""
         with tempfile.TemporaryDirectory() as tmpdir:
-            os.chdir(tmpdir)
+            # Save original directory to avoid Windows file deletion issues
+            original_cwd = os.getcwd()
+            try:
+                os.chdir(tmpdir)
 
-            with open("input.txt", "w") as f:
-                f.write("x = ${x}\n")
+                with open("input.txt", "w") as f:
+                    f.write("x = ${x}\n")
 
-            calc_script = Path(tmpdir) / "calc.py"
-            with open(calc_script, "w") as f:
-                f.write("""#!/usr/bin/env python3
+                calc_script = Path(tmpdir) / "calc.py"
+                with open(calc_script, "w") as f:
+                    f.write("""#!/usr/bin/env python3
 import json
 
 with open('input.txt', 'r') as f:
@@ -426,33 +450,36 @@ def test_fzr_with_multiple_dict_outputs(self):
     f.write(f"stats={json.dumps(stats)}\\n")
     f.write(f"meta={json.dumps(meta)}\\n")
 """)
-            os.chmod(calc_script, 0o755)
-
-            model = {
-                "varprefix": "$",
-                "delim": "{}",
-                "output": {
-                    "stats": "grep 'stats=' output.txt | cut -d'=' -f2",
-                    "meta": "grep 'meta=' output.txt | cut -d'=' -f2"
+                os.chmod(calc_script, 0o755)
+
+                model = {
+                    "varprefix": "$",
+                    "delim": "{}",
+                    "output": {
+                        "stats": "grep 'stats=' output.txt | cut -d'=' -f2",
+                        "meta": "grep 'meta=' output.txt | cut -d'=' -f2"
+                    }
                 }
-            }
 
-            results = fz.fzr(
-                input_path="input.txt",
-                input_variables={"x": [5, 10]},
-                model=model,
-                calculators=f"sh://python3 {calc_script}"
-            )
-
-            # Check both dicts flattened
-            assert 'stats_min' in results.columns
-            assert 'stats_max' in results.columns
-            assert 'meta_name' in results.columns
-            assert 'meta_id' in results.columns
-
-            # Verify values
-            assert results['meta_name'].iloc[0] == 'case5'
-            assert results['meta_id'].iloc[0] == 500
+                results = fz.fzr(
+                    input_path="input.txt",
+                    input_variables={"x": [5, 10]},
+                    model=model,
+                    calculators=f"sh://python3 {calc_script}"
+                )
+
+                # Check both dicts flattened
+                assert 'stats_min' in results.columns
+                assert 'stats_max' in results.columns
+                assert 'meta_name' in results.columns
+                assert 'meta_id' in results.columns
+
+                # Verify values
+                assert results['meta_name'].iloc[0] == 'case5'
+                assert results['meta_id'].iloc[0] == 500
+            finally:
+                # Restore original directory to allow cleanup on Windows
+                os.chdir(original_cwd)
 
 
 class TestEdgeCases: