Scale dense charts with result count

src/charts.py

@@ -24,6 +24,13 @@ PlotBackground = Literal["Dark", "White"]
 DEFAULT_PALETTE: PaletteName = "Citrus"
 DEFAULT_BACKGROUND: PlotBackground = "Dark"
 
+RANKING_MIN_HEIGHT_PX = 680
+RANKING_ROW_HEIGHT_PX = 44
+RANKING_VERTICAL_PADDING_PX = 280
+MATRIX_MIN_HEIGHT_PX = 720
+MATRIX_ROW_HEIGHT_PX = 36
+MATRIX_VERTICAL_PADDING_PX = 260
+
 # Separate categorical palettes for each grouping dimension.
 # Model and harness colors intentionally start from different hue families so
 # switching "Color by" remains visually obvious.
@@ -260,8 +267,9 @@ def apply_plot_theme(fig: Figure, background_name: str | None = DEFAULT_BACKGROU
             "itemdoubleclick": False,
         },
     )
-    # Remove fixed dimensions so Gradio/Plotly can size to the browser/container.
-    fig.update_layout(width=None, height=None)
+    # Width remains responsive. Preserve any explicit height set by dense
+    # categorical charts so Plotly has enough vertical room for every label.
+    fig.update_layout(width=None)
     fig.update_xaxes(
         automargin=True,
         color=theme["text_muted"],
@@ -651,6 +659,17 @@ def create_token_pareto_frontier_plot(
     )
 
 
+def _categorical_chart_height(
+    n_rows: int,
+    *,
+    min_height: int,
+    row_height: int,
+    vertical_padding: int,
+) -> int:
+    """Scale dense categorical charts vertically so labels remain readable."""
+    return max(min_height, row_height * max(n_rows, 0) + vertical_padding)
+
+
 def create_ranking_plot(
     dataframe: pd.DataFrame,
     metric_column: str,
@@ -726,9 +745,18 @@ def create_ranking_plot(
             "autorange": "reversed",
             "categoryorder": "array",
             "categoryarray": agent_order,
+            "tickmode": "array",
+            "tickvals": agent_order,
+            "ticktext": agent_order,
         },
         legend_title_text=color_by,
         barmode="group",
+        height=_categorical_chart_height(
+            len(agent_order),
+            min_height=RANKING_MIN_HEIGHT_PX,
+            row_height=RANKING_ROW_HEIGHT_PX,
+            vertical_padding=RANKING_VERTICAL_PADDING_PX,
+        ),
     )
     return apply_plot_theme(fig, background_name)
 
@@ -877,11 +905,23 @@ def create_matrix_plot(
             hoverongaps=False,
         )
     )
+    row_labels = [str(value) for value in matrix.index]
     fig.update_layout(
         title=title,
         xaxis={"title": "Benchmark"},
-        yaxis={"title": "Model / Harness", "autorange": "reversed"},
-        height=max(480, 28 * len(matrix.index) + 180),
+        yaxis={
+            "title": "Model / Harness",
+            "autorange": "reversed",
+            "tickmode": "array",
+            "tickvals": row_labels,
+            "ticktext": row_labels,
+        },
+        height=_categorical_chart_height(
+            len(row_labels),
+            min_height=MATRIX_MIN_HEIGHT_PX,
+            row_height=MATRIX_ROW_HEIGHT_PX,
+            vertical_padding=MATRIX_VERTICAL_PADDING_PX,
+        ),
     )
     return apply_plot_theme(fig, background_name)
 
@@ -910,10 +950,22 @@ def create_coverage_matrix_plot(
             hovertemplate="Agent: %{y}<br>Benchmark: %{x}<br>Status: %{text}<extra></extra>",
         )
     )
+    row_labels = [str(value) for value in display.index]
     fig.update_layout(
         title="Benchmark coverage",
         xaxis={"title": "Benchmark"},
-        yaxis={"title": "Model / Harness", "autorange": "reversed"},
-        height=max(480, 28 * len(display.index) + 180),
+        yaxis={
+            "title": "Model / Harness",
+            "autorange": "reversed",
+            "tickmode": "array",
+            "tickvals": row_labels,
+            "ticktext": row_labels,
+        },
+        height=_categorical_chart_height(
+            len(row_labels),
+            min_height=MATRIX_MIN_HEIGHT_PX,
+            row_height=MATRIX_ROW_HEIGHT_PX,
+            vertical_padding=MATRIX_VERTICAL_PADDING_PX,
+        ),
     )
     return apply_plot_theme(fig, background_name)

tests/test_pr2_analytics.py

@@ -203,3 +203,18 @@ def test_ranking_plot_order_can_be_value_or_alphabetical():
     assert list(largest.layout.yaxis.categoryarray) == ["c / h", "b / h", "a / h"]
     assert list(lowest.layout.yaxis.categoryarray) == ["a / h", "b / h", "c / h"]
     assert list(alpha.layout.yaxis.categoryarray) == ["a / h", "b / h", "c / h"]
+
+
+def test_matrix_height_scales_with_rows_and_keeps_all_y_labels():
+    rows = 18
+    matrix = pd.DataFrame(
+        {"bench-a": range(rows), "bench-b": range(rows)},
+        index=[f"model-{i} / harness" for i in range(rows)],
+    )
+
+    figure = create_matrix_plot(matrix, "Dense matrix", "Score (%)")
+
+    assert figure.layout.height >= 900
+    assert figure.layout.yaxis.tickmode == "array"
+    assert len(figure.layout.yaxis.tickvals) == rows
+    assert len(figure.layout.yaxis.ticktext) == rows

tests/test_ranking_refinements.py

@@ -71,3 +71,22 @@ def test_rankings_page_places_paired_comparisons_first_and_has_no_score_all_benc
     assert 'ranking_sections = [("Score", "Score", BENCHMARK_NAMES, DEFAULT_BENCHMARK)]' in rankings
     assert '"Best first"' in rankings
     assert '"Best last"' in rankings
+
+
+def test_ranking_plot_height_scales_with_agents_and_keeps_all_y_labels():
+    rows = 16
+    df = pd.DataFrame(
+        {
+            "Model": [f"model-{i}" for i in range(rows)],
+            "Harness": ["h"] * rows,
+            "Benchmark": ["bench"] * rows,
+            "Score (%)": list(range(rows)),
+        }
+    )
+
+    figure = create_ranking_plot(df, "Score (%)", "Score", True, sort_order="Best first")
+
+    assert figure.layout.height >= 900
+    assert figure.layout.yaxis.tickmode == "array"
+    assert len(figure.layout.yaxis.tickvals) == rows
+    assert len(figure.layout.yaxis.ticktext) == rows