Skip to content

~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.

31.1 KB
Raw Blame
"""
Tests for benchmarks/comparison/evaluator.py

Tests cover:
- compare_configurations function
- _calculate_average_metrics function
- _evaluate_single_configuration function structure
- Visualization creation functions
"""

from unittest.mock import patch
import pytest


class TestCompareConfigurationsValidation:
    """Tests for compare_configurations input validation."""

    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_returns_error_for_empty_configurations(
        self, mock_makedirs, tmp_path
    ):
        """Test that empty configurations list returns error."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        result = compare_configurations(
            query="test query",
            configurations=[],
            output_dir=str(tmp_path),
        )

        assert "error" in result
        assert result["error"] == "No configurations provided"

    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_creates_output_directory(self, mock_makedirs):
        """Test that output directory is created."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        # Empty configs will return early with error but still create dir
        compare_configurations(
            query="test",
            configurations=[],
            output_dir="/tmp/test_output",
        )

        mock_makedirs.assert_called_with("/tmp/test_output", exist_ok=True)


class TestCalculateAverageMetrics:
    """Tests for _calculate_average_metrics function."""

    def test_calculates_average_quality_metrics(self):
        """Test averaging of quality metrics."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _calculate_average_metrics,
        )

        results = [
            {"quality_metrics": {"accuracy": 0.8, "source_count": 10}},
            {"quality_metrics": {"accuracy": 0.6, "source_count": 8}},
        ]

        avg = _calculate_average_metrics(results)

        assert avg["quality_metrics"]["accuracy"] == pytest.approx(0.7)
        assert avg["quality_metrics"]["source_count"] == pytest.approx(9.0)

    def test_calculates_average_speed_metrics(self):
        """Test averaging of speed metrics."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _calculate_average_metrics,
        )

        results = [
            {
                "speed_metrics": {
                    "total_duration": 100,
                    "duration_per_question": 10,
                }
            },
            {
                "speed_metrics": {
                    "total_duration": 80,
                    "duration_per_question": 8,
                }
            },
        ]

        avg = _calculate_average_metrics(results)

        assert avg["speed_metrics"]["total_duration"] == pytest.approx(90.0)
        assert avg["speed_metrics"]["duration_per_question"] == pytest.approx(
            9.0
        )

    def test_calculates_average_resource_metrics(self):
        """Test averaging of resource metrics."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _calculate_average_metrics,
        )

        results = [
            {"resource_metrics": {"memory_mb": 100}},
            {"resource_metrics": {"memory_mb": 200}},
        ]

        avg = _calculate_average_metrics(results)

        assert avg["resource_metrics"]["memory_mb"] == pytest.approx(150.0)

    def test_handles_empty_results(self):
        """Test handling of empty results list."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _calculate_average_metrics,
        )

        avg = _calculate_average_metrics([])

        assert avg == {}

    def test_handles_missing_metrics_categories(self):
        """Test handling of results with missing metric categories."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _calculate_average_metrics,
        )

        results = [
            {"quality_metrics": {"accuracy": 0.8}},
            {},  # No metrics
        ]

        avg = _calculate_average_metrics(results)

        # Should still have quality_metrics
        assert "quality_metrics" in avg
        assert avg["quality_metrics"]["accuracy"] == pytest.approx(0.8)

    def test_handles_none_values(self):
        """Test handling of None values in metrics."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _calculate_average_metrics,
        )

        results = [
            {"quality_metrics": {"accuracy": 0.8, "optional": None}},
            {"quality_metrics": {"accuracy": 0.6, "optional": 0.5}},
        ]

        avg = _calculate_average_metrics(results)

        # None values should be filtered out
        assert avg["quality_metrics"]["accuracy"] == pytest.approx(0.7)
        # Optional should only have one valid value
        assert avg["quality_metrics"]["optional"] == pytest.approx(0.5)

    def test_handles_mixed_metric_keys(self):
        """Test handling when results have different metric keys."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _calculate_average_metrics,
        )

        results = [
            {"quality_metrics": {"metric_a": 0.8}},
            {"quality_metrics": {"metric_b": 0.6}},
        ]

        avg = _calculate_average_metrics(results)

        # Both metrics should be present
        assert avg["quality_metrics"]["metric_a"] == pytest.approx(0.8)
        assert avg["quality_metrics"]["metric_b"] == pytest.approx(0.6)


class TestMetricWeightsDefault:
    """Tests for default metric weights."""

    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_default_metric_weights(self, mock_makedirs):
        """Test that default metric weights are applied."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        compare_configurations(
            query="test",
            configurations=[],
            output_dir="/tmp/test",
        )

        # Empty configs returns early, but we can verify the function signature
        assert callable(compare_configurations)


class TestVisualizationFunctions:
    """Tests for visualization helper functions."""

    def test_create_metric_comparison_chart_exists(self):
        """Test that _create_metric_comparison_chart function exists."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _create_metric_comparison_chart,
        )

        assert callable(_create_metric_comparison_chart)

    def test_create_spider_chart_exists(self):
        """Test that _create_spider_chart function exists."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _create_spider_chart,
        )

        assert callable(_create_spider_chart)

    def test_create_pareto_chart_exists(self):
        """Test that _create_pareto_chart function exists."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _create_pareto_chart,
        )

        assert callable(_create_pareto_chart)

    def test_create_comparison_visualizations_exists(self):
        """Test that _create_comparison_visualizations function exists."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _create_comparison_visualizations,
        )

        assert callable(_create_comparison_visualizations)


class TestEvaluateSingleConfiguration:
    """Tests for _evaluate_single_configuration function."""

    def test_evaluate_single_configuration_exists(self):
        """Test that _evaluate_single_configuration function exists."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _evaluate_single_configuration,
        )

        assert callable(_evaluate_single_configuration)

    def test_function_signature(self):
        """Test function has expected parameters."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _evaluate_single_configuration,
        )
        import inspect

        sig = inspect.signature(_evaluate_single_configuration)
        params = list(sig.parameters.keys())

        assert "query" in params
        assert "config" in params


class TestConfigurationResultStructure:
    """Tests for expected result structure."""

    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_compare_returns_dict(self, mock_makedirs):
        """Test that compare_configurations returns a dictionary."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        result = compare_configurations(
            query="test",
            configurations=[],
            output_dir="/tmp/test",
        )

        assert isinstance(result, dict)

    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_error_result_has_error_key(self, mock_makedirs):
        """Test that error results have 'error' key."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        result = compare_configurations(
            query="test",
            configurations=[],
            output_dir="/tmp/test",
        )

        assert "error" in result


class TestCompareConfigurationsWithMocks:
    """Tests for compare_configurations with full mocking."""

    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._evaluate_single_configuration"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._create_comparison_visualizations"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_combined_score"
    )
    @patch(
        "local_deep_research.security.file_write_verifier.write_json_verified"
    )
    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_compare_single_configuration(
        self, mock_makedirs, mock_write, mock_combined, mock_viz, mock_eval
    ):
        """Test comparing a single configuration."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        mock_combined.return_value = 0.8
        mock_eval.return_value = {
            "success": True,
            "quality_metrics": {"overall_quality": 0.8},
            "speed_metrics": {"total_duration": 10.0},
            "resource_metrics": {},
        }

        result = compare_configurations(
            query="test query",
            configurations=[{"name": "Config 1", "iterations": 2}],
            output_dir="/tmp/test",
            repetitions=1,
        )

        assert result["configurations_tested"] == 1
        assert result["successful_configurations"] == 1
        assert len(result["results"]) == 1

    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._evaluate_single_configuration"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._create_comparison_visualizations"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_combined_score"
    )
    @patch(
        "local_deep_research.security.file_write_verifier.write_json_verified"
    )
    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_compare_multiple_configurations(
        self, mock_makedirs, mock_write, mock_combined, mock_viz, mock_eval
    ):
        """Test comparing multiple configurations."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        mock_combined.return_value = 0.8
        mock_eval.side_effect = [
            {
                "success": True,
                "quality_metrics": {"overall_quality": 0.8},
                "speed_metrics": {"total_duration": 10.0},
                "resource_metrics": {},
            },
            {
                "success": True,
                "quality_metrics": {"overall_quality": 0.7},
                "speed_metrics": {"total_duration": 15.0},
                "resource_metrics": {},
            },
        ]

        result = compare_configurations(
            query="test",
            configurations=[
                {"name": "Config 1"},
                {"name": "Config 2"},
            ],
            output_dir="/tmp/test",
        )

        assert result["configurations_tested"] == 2
        assert result["successful_configurations"] == 2

    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._evaluate_single_configuration"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._create_comparison_visualizations"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_combined_score"
    )
    @patch(
        "local_deep_research.security.file_write_verifier.write_json_verified"
    )
    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_compare_handles_failed_configuration(
        self, mock_makedirs, mock_write, mock_combined, mock_viz, mock_eval
    ):
        """Test handling of failed configuration."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        mock_combined.return_value = 0.0
        mock_eval.return_value = {
            "success": False,
            "error": "Config failed",
        }

        result = compare_configurations(
            query="test",
            configurations=[{"name": "Failing Config"}],
            output_dir="/tmp/test",
        )

        assert result["failed_configurations"] == 1
        assert result["successful_configurations"] == 0

    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._evaluate_single_configuration"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._create_comparison_visualizations"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_combined_score"
    )
    @patch(
        "local_deep_research.security.file_write_verifier.write_json_verified"
    )
    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_compare_with_multiple_repetitions(
        self, mock_makedirs, mock_write, mock_combined, mock_viz, mock_eval
    ):
        """Test compare with multiple repetitions per configuration."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        mock_combined.return_value = 0.8
        # Three repetitions for one config
        mock_eval.side_effect = [
            {
                "success": True,
                "quality_metrics": {"overall_quality": 0.8},
                "speed_metrics": {"total_duration": 10.0},
                "resource_metrics": {},
            },
            {
                "success": True,
                "quality_metrics": {"overall_quality": 0.85},
                "speed_metrics": {"total_duration": 9.0},
                "resource_metrics": {},
            },
            {
                "success": True,
                "quality_metrics": {"overall_quality": 0.75},
                "speed_metrics": {"total_duration": 11.0},
                "resource_metrics": {},
            },
        ]

        result = compare_configurations(
            query="test",
            configurations=[{"name": "Config 1"}],
            output_dir="/tmp/test",
            repetitions=3,
        )

        assert result["repetitions"] == 3
        assert result["results"][0]["runs_completed"] == 3

    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._evaluate_single_configuration"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._create_comparison_visualizations"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_combined_score"
    )
    @patch(
        "local_deep_research.security.file_write_verifier.write_json_verified"
    )
    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_compare_custom_metric_weights(
        self, mock_makedirs, mock_write, mock_combined, mock_viz, mock_eval
    ):
        """Test compare with custom metric weights."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        mock_combined.return_value = 0.8
        mock_eval.return_value = {
            "success": True,
            "quality_metrics": {"overall_quality": 0.8},
            "speed_metrics": {"total_duration": 10.0},
            "resource_metrics": {},
        }

        custom_weights = {"quality": 0.8, "speed": 0.2, "resource": 0.0}

        result = compare_configurations(
            query="test",
            configurations=[{"name": "Config 1"}],
            output_dir="/tmp/test",
            metric_weights=custom_weights,
        )

        assert result["metric_weights"] == custom_weights

    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._evaluate_single_configuration"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator._create_comparison_visualizations"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_combined_score"
    )
    @patch(
        "local_deep_research.security.file_write_verifier.write_json_verified"
    )
    @patch("local_deep_research.benchmarks.comparison.evaluator.os.makedirs")
    def test_results_sorted_by_score_descending(
        self, mock_makedirs, mock_write, mock_combined, mock_viz, mock_eval
    ):
        """Test that results are sorted by score in descending order."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            compare_configurations,
        )

        mock_combined.side_effect = [0.5, 0.9, 0.7]
        mock_eval.side_effect = [
            {
                "success": True,
                "quality_metrics": {"overall_quality": 0.5},
                "speed_metrics": {"total_duration": 10.0},
                "resource_metrics": {},
            },
            {
                "success": True,
                "quality_metrics": {"overall_quality": 0.9},
                "speed_metrics": {"total_duration": 10.0},
                "resource_metrics": {},
            },
            {
                "success": True,
                "quality_metrics": {"overall_quality": 0.7},
                "speed_metrics": {"total_duration": 10.0},
                "resource_metrics": {},
            },
        ]

        result = compare_configurations(
            query="test",
            configurations=[
                {"name": "Low"},
                {"name": "High"},
                {"name": "Mid"},
            ],
            output_dir="/tmp/test",
        )

        # Successful results should be sorted by score descending
        successful = [r for r in result["results"] if r.get("success")]
        scores = [r.get("overall_score", 0) for r in successful]
        assert scores == sorted(scores, reverse=True)


class TestEvaluateSingleConfigurationFull:
    """Full tests for _evaluate_single_configuration function."""

    from unittest.mock import Mock

    @patch("local_deep_research.benchmarks.comparison.evaluator.get_llm")
    @patch("local_deep_research.benchmarks.comparison.evaluator.get_search")
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.AdvancedSearchSystem"
    )
    @patch("local_deep_research.benchmarks.comparison.evaluator.SpeedProfiler")
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.ResourceMonitor"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_quality_metrics"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_speed_metrics"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_resource_metrics"
    )
    def test_successful_evaluation(
        self,
        mock_res_metrics,
        mock_speed_metrics,
        mock_quality_metrics,
        mock_res_monitor,
        mock_profiler,
        mock_search_system,
        mock_get_search,
        mock_get_llm,
    ):
        """Test successful configuration evaluation."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _evaluate_single_configuration,
        )
        from unittest.mock import Mock

        # Setup mocks
        mock_llm = Mock()
        mock_get_llm.return_value = mock_llm

        mock_search = Mock()
        mock_get_search.return_value = mock_search

        mock_system = Mock()
        mock_system.analyze_topic.return_value = {
            "findings": [{"phase": 1, "content": "test"}],
            "current_knowledge": "Test knowledge",
        }
        mock_system.all_links_of_system = ["http://example.com"]
        mock_search_system.return_value = mock_system

        mock_profiler_instance = Mock()
        mock_profiler_instance.timer.return_value.__enter__ = Mock()
        mock_profiler_instance.timer.return_value.__exit__ = Mock(
            return_value=False
        )
        mock_profiler_instance.get_summary.return_value = {}
        mock_profiler_instance.get_timings.return_value = {}
        mock_profiler.return_value = mock_profiler_instance

        mock_res_monitor_instance = Mock()
        mock_res_monitor_instance.get_combined_stats.return_value = {}
        mock_res_monitor.return_value = mock_res_monitor_instance

        mock_quality_metrics.return_value = {"overall_quality": 0.8}
        mock_speed_metrics.return_value = {"total_duration": 10.0}
        mock_res_metrics.return_value = {}

        config = {"iterations": 2, "search_strategy": "iterdrag"}

        result = _evaluate_single_configuration(
            query="test query",
            config=config,
        )

        assert result["success"] is True
        assert "quality_metrics" in result
        assert "speed_metrics" in result

    @patch("local_deep_research.benchmarks.comparison.evaluator.get_llm")
    @patch("local_deep_research.benchmarks.comparison.evaluator.SpeedProfiler")
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.ResourceMonitor"
    )
    def test_evaluation_handles_llm_error(
        self,
        mock_res_monitor,
        mock_profiler,
        mock_get_llm,
    ):
        """Test that evaluation handles LLM initialization errors."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _evaluate_single_configuration,
        )
        from unittest.mock import Mock

        mock_get_llm.side_effect = Exception("LLM init failed")

        mock_profiler_instance = Mock()
        mock_profiler_instance.timer.return_value.__enter__ = Mock()
        mock_profiler_instance.timer.return_value.__exit__ = Mock(
            return_value=False
        )
        mock_profiler_instance.get_timings.return_value = {}
        mock_profiler.return_value = mock_profiler_instance

        mock_res_monitor_instance = Mock()
        mock_res_monitor_instance.get_combined_stats.return_value = {}
        mock_res_monitor.return_value = mock_res_monitor_instance

        config = {"iterations": 2}

        result = _evaluate_single_configuration(
            query="test",
            config=config,
        )

        assert result["success"] is False
        assert "error" in result

    @patch("local_deep_research.benchmarks.comparison.evaluator.get_llm")
    @patch("local_deep_research.benchmarks.comparison.evaluator.get_search")
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.AdvancedSearchSystem"
    )
    @patch("local_deep_research.benchmarks.comparison.evaluator.SpeedProfiler")
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.ResourceMonitor"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_quality_metrics"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_speed_metrics"
    )
    @patch(
        "local_deep_research.benchmarks.comparison.evaluator.calculate_resource_metrics"
    )
    def test_evaluation_uses_config_parameters(
        self,
        mock_res_metrics,
        mock_speed_metrics,
        mock_quality_metrics,
        mock_res_monitor,
        mock_profiler,
        mock_search_system,
        mock_get_search,
        mock_get_llm,
    ):
        """Test that configuration parameters are applied correctly."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _evaluate_single_configuration,
        )
        from unittest.mock import Mock

        mock_llm = Mock()
        mock_get_llm.return_value = mock_llm

        mock_search = Mock()
        mock_get_search.return_value = mock_search

        mock_system = Mock()
        mock_system.analyze_topic.return_value = {
            "findings": [],
            "current_knowledge": "",
        }
        mock_search_system.return_value = mock_system

        mock_profiler_instance = Mock()
        mock_profiler_instance.timer.return_value.__enter__ = Mock()
        mock_profiler_instance.timer.return_value.__exit__ = Mock(
            return_value=False
        )
        mock_profiler_instance.get_summary.return_value = {}
        mock_profiler_instance.get_timings.return_value = {}
        mock_profiler.return_value = mock_profiler_instance

        mock_res_monitor_instance = Mock()
        mock_res_monitor_instance.get_combined_stats.return_value = {}
        mock_res_monitor.return_value = mock_res_monitor_instance

        mock_quality_metrics.return_value = {}
        mock_speed_metrics.return_value = {}
        mock_res_metrics.return_value = {}

        config = {
            "iterations": 5,
            "questions_per_iteration": 4,
            "search_strategy": "focused_iteration",
        }

        _evaluate_single_configuration(
            query="test",
            config=config,
        )

        # Verify system was configured with our parameters
        call_kwargs = mock_search_system.call_args[1]
        assert call_kwargs["max_iterations"] == 5
        assert call_kwargs["questions_per_iteration"] == 4
        assert call_kwargs["strategy_name"] == "focused_iteration"


class TestVisualizationCreation:
    """Tests for visualization creation functions."""

    from unittest.mock import Mock

    @patch("local_deep_research.benchmarks.comparison.evaluator.plt")
    def test_create_comparison_visualizations_no_successful(self, mock_plt):
        """Test visualization with no successful results."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _create_comparison_visualizations,
        )

        report = {"results": [{"success": False}]}

        # Should not raise
        _create_comparison_visualizations(
            report, output_dir="/tmp/test", timestamp="20240101"
        )

    @patch("local_deep_research.benchmarks.comparison.evaluator.plt")
    def test_create_metric_comparison_chart_single_metric(self, mock_plt):
        """Test metric comparison chart with single metric."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _create_metric_comparison_chart,
        )
        from unittest.mock import Mock, MagicMock

        mock_fig = Mock()
        mock_ax = MagicMock()
        mock_plt.subplots.return_value = (mock_fig, mock_ax)

        # barh returns a BarContainer which is iterable; mock it with a list
        # of mock bars that support get_width/get_y/get_height
        mock_bar = Mock()
        mock_bar.get_width.return_value = 0.8
        mock_bar.get_y.return_value = 0.0
        mock_bar.get_height.return_value = 0.5
        mock_ax.barh.return_value = [mock_bar]

        results = [
            {
                "name": "Config 1",
                "avg_metrics": {"quality_metrics": {"overall_quality": 0.8}},
            }
        ]

        _create_metric_comparison_chart(
            results,
            ["Config 1"],
            ["overall_quality"],
            "quality_metrics",
            "Test",
            "/tmp/test.png",
        )

        mock_plt.savefig.assert_called()

    @patch("local_deep_research.benchmarks.comparison.evaluator.plt")
    def test_create_pareto_chart_with_data(self, mock_plt):
        """Test pareto chart creation with data."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _create_pareto_chart,
        )
        from unittest.mock import Mock

        mock_fig = Mock()
        mock_ax = Mock()
        mock_plt.subplots.return_value = (mock_fig, mock_ax)
        mock_plt.figure.return_value = mock_fig

        results = [
            {
                "name": "Config 1",
                "avg_metrics": {
                    "quality_metrics": {"overall_quality": 0.8},
                    "speed_metrics": {"total_duration": 10.0},
                },
            },
            {
                "name": "Config 2",
                "avg_metrics": {
                    "quality_metrics": {"overall_quality": 0.6},
                    "speed_metrics": {"total_duration": 5.0},
                },
            },
        ]

        _create_pareto_chart(results, "/tmp/pareto.png")

        mock_plt.savefig.assert_called()

    @patch("local_deep_research.benchmarks.comparison.evaluator.plt")
    def test_create_comparison_visualizations_creates_files(self, mock_plt):
        """Test that visualizations create output files."""
        from local_deep_research.benchmarks.comparison.evaluator import (
            _create_comparison_visualizations,
        )
        from unittest.mock import Mock, MagicMock

        def _make_mock_ax():
            """Create a mock ax that supports barh returning iterable bars."""
            ax = MagicMock()
            bar = Mock()
            bar.get_width.return_value = 0.5
            bar.get_y.return_value = 0.0
            bar.get_height.return_value = 0.5
            ax.barh.return_value = [bar]
            return ax

        mock_fig = Mock()

        def _subplots_side_effect(*args, **kwargs):
            """Return (fig, axes) with correct number of axes."""
            nrows = args[0] if args else kwargs.get("nrows", 1)
            if nrows == 1:
                return (mock_fig, _make_mock_ax())
            return (mock_fig, [_make_mock_ax() for _ in range(nrows)])

        mock_plt.subplots.side_effect = _subplots_side_effect
        mock_plt.figure.return_value = mock_fig

        report = {
            "results": [
                {
                    "name": "Config 1",
                    "success": True,
                    "overall_score": 0.8,
                    "avg_metrics": {
                        "quality_metrics": {"overall_quality": 0.8},
                        "speed_metrics": {"total_duration": 10.0},
                        "resource_metrics": {},
                    },
                }
            ]
        }

        _create_comparison_visualizations(
            report, output_dir="/tmp/test", timestamp="20240101"
        )

        # Should have called savefig multiple times
        assert mock_plt.savefig.called