scitex 2.7.3__py3-none-any.whl → 2.10.0__py3-none-any.whl

scitex/fts/_fig/_editor/_gui/_flask_editor/_renderer.py

@@ -0,0 +1,853 @@
+#!/usr/bin/env python3
+# File: ./src/scitex/vis/editor/flask_editor/renderer.py
+"""Figure rendering for Flask editor - supports single and multi-axis figures."""
+
+import base64
+import io
+from typing import Any, Dict, Optional, Tuple
+
+import matplotlib
+
+matplotlib.use("Agg")
+import matplotlib.pyplot as plt
+import numpy as np
+from matplotlib.ticker import MaxNLocator
+from PIL import Image
+
+from scitex.plt.styles import get_default_dpi
+
+from ._bbox import extract_bboxes, extract_bboxes_multi
+from ._plotter import plot_from_csv, plot_from_recipe
+
+# mm to pt conversion factor
+MM_TO_PT = 2.83465
+
+
+def render_preview_with_bboxes(
+    csv_data,
+    overrides: Dict[str, Any],
+    axis_fontsize: int = 7,
+    metadata: Optional[Dict[str, Any]] = None,
+    dark_mode: bool = False,
+) -> Tuple[str, Dict[str, Any], Dict[str, int]]:
+    """Render figure and return base64 PNG along with element bounding boxes.
+
+    Args:
+        csv_data: DataFrame containing CSV data
+        overrides: Dictionary with override settings
+        axis_fontsize: Default font size for axis labels
+        metadata: Optional JSON metadata (new schema with axes dict)
+        dark_mode: Whether to render with dark mode colors (light text/spines)
+
+    Returns:
+        tuple: (base64_image_data, bboxes_dict, image_size)
+    """
+    # Check if this is a multi-axis figure (new schema)
+    if metadata and "axes" in metadata and isinstance(metadata.get("axes"), dict):
+        return render_multi_axis_preview(csv_data, overrides, metadata, dark_mode)
+
+    # Fall back to single-axis rendering
+    return render_single_axis_preview(csv_data, overrides, axis_fontsize, dark_mode)
+
+
+def render_single_axis_preview(
+    csv_data,
+    overrides: Dict[str, Any],
+    axis_fontsize: int = 7,
+    dark_mode: bool = False,
+) -> Tuple[str, Dict[str, Any], Dict[str, int]]:
+    """Render single-axis figure (legacy mode)."""
+    o = overrides
+
+    # Dimensions
+    dpi = o.get("dpi", get_default_dpi())
+    fig_size = o.get("fig_size", [3.15, 2.68])
+
+    # Font sizes
+    axis_fontsize = o.get("axis_fontsize", 7)
+    tick_fontsize = o.get("tick_fontsize", 7)
+    title_fontsize = o.get("title_fontsize", 8)
+
+    # Line/axis thickness
+    linewidth_pt = o.get("linewidth", 0.57)
+    axis_width_pt = o.get("axis_width", 0.2) * MM_TO_PT
+    tick_length_pt = o.get("tick_length", 0.8) * MM_TO_PT
+    tick_width_pt = o.get("tick_width", 0.2) * MM_TO_PT
+    tick_direction = o.get("tick_direction", "out")
+    x_n_ticks = o.get("x_n_ticks", o.get("n_ticks", 4))
+    y_n_ticks = o.get("y_n_ticks", o.get("n_ticks", 4))
+    hide_x_ticks = o.get("hide_x_ticks", False)
+    hide_y_ticks = o.get("hide_y_ticks", False)
+
+    transparent = o.get("transparent", True)
+
+    # Create figure
+    fig, ax = plt.subplots(figsize=fig_size, dpi=dpi)
+    _apply_background(fig, ax, o, transparent)
+
+    # Plot from CSV data
+    if csv_data is not None:
+        plot_from_csv(ax, csv_data, overrides, linewidth=linewidth_pt)
+    else:
+        ax.text(
+            0.5,
+            0.5,
+            "No plot data available\n(CSV not found)",
+            ha="center",
+            va="center",
+            transform=ax.transAxes,
+            fontsize=axis_fontsize,
+        )
+
+    # Apply labels
+    _apply_labels(ax, o, title_fontsize, axis_fontsize)
+
+    # Tick styling
+    _apply_tick_styling(
+        ax,
+        tick_fontsize,
+        tick_length_pt,
+        tick_width_pt,
+        tick_direction,
+        x_n_ticks,
+        y_n_ticks,
+        hide_x_ticks,
+        hide_y_ticks,
+    )
+
+    # Apply grid, limits, spines
+    _apply_style(ax, o, axis_width_pt)
+
+    # Apply annotations
+    _apply_annotations(ax, o, axis_fontsize)
+
+    # Apply caption (below figure)
+    caption_artist = _apply_caption(fig, o)
+
+    # Apply dark mode styling if requested
+    if dark_mode:
+        _apply_dark_theme(ax)
+        # Also style caption if present
+        if caption_artist:
+            caption_artist.set_color(DARK_THEME_TEXT_COLOR)
+
+    fig.tight_layout()
+
+    # Get element bounding boxes BEFORE saving (need renderer)
+    fig.canvas.draw()
+    renderer = fig.canvas.get_renderer()
+
+    # Save to buffer first to get actual image size
+    buf = io.BytesIO()
+    fig.savefig(
+        buf, format="png", dpi=dpi, bbox_inches="tight", transparent=transparent
+    )
+    buf.seek(0)
+
+    # Get actual saved image dimensions
+    img = Image.open(buf)
+    img_width, img_height = img.size
+    buf.seek(0)
+
+    # Get bboxes
+    bboxes = extract_bboxes(fig, ax, renderer, img_width, img_height)
+
+    img_data = base64.b64encode(buf.read()).decode("utf-8")
+    plt.close(fig)
+
+    return img_data, bboxes, {"width": img_width, "height": img_height}
+
+
+def render_multi_axis_preview(
+    csv_data,
+    overrides: Dict[str, Any],
+    metadata: Dict[str, Any],
+    dark_mode: bool = False,
+) -> Tuple[str, Dict[str, Any], Dict[str, int]]:
+    """Render multi-axis figure from new schema (scitex.plt.figure.recipe).
+
+    Args:
+        csv_data: DataFrame containing CSV data
+        overrides: Dictionary with override settings
+        metadata: JSON metadata with axes dict
+        dark_mode: Whether to render with dark mode colors
+
+    Returns:
+        tuple: (base64_image_data, bboxes_dict, image_size)
+    """
+    o = overrides
+    axes_spec = metadata.get("axes", {})
+    fig_spec = metadata.get("figure", {})
+
+    # Get grid dimensions from axes positions
+    nrows, ncols = _get_grid_dimensions(axes_spec)
+
+    # Figure dimensions
+    dpi = fig_spec.get("dpi", o.get("dpi", get_default_dpi()))
+    size_mm = fig_spec.get("size_mm", [176, 106])
+    # Convert mm to inches (1 inch = 25.4 mm)
+    fig_size = (size_mm[0] / 25.4, size_mm[1] / 25.4)
+
+    # Font sizes (from overrides)
+    axis_fontsize = o.get("axis_fontsize", 7)
+    tick_fontsize = o.get("tick_fontsize", 7)
+    title_fontsize = o.get("title_fontsize", 8)
+
+    # Line/axis thickness
+    linewidth_pt = o.get("linewidth", 0.57)
+    axis_width_pt = o.get("axis_width", 0.2) * MM_TO_PT
+    tick_length_pt = o.get("tick_length", 0.8) * MM_TO_PT
+    tick_width_pt = o.get("tick_width", 0.2) * MM_TO_PT
+    tick_direction = o.get("tick_direction", "out")
+    x_n_ticks = o.get("x_n_ticks", o.get("n_ticks", 4))
+    y_n_ticks = o.get("y_n_ticks", o.get("n_ticks", 4))
+
+    transparent = o.get("transparent", True)
+
+    # Create multi-axis figure
+    fig, axes_array = plt.subplots(nrows, ncols, figsize=fig_size, dpi=dpi)
+
+    # Handle 1D or 2D array
+    if nrows == 1 and ncols == 1:
+        axes_array = np.array([[axes_array]])
+    elif nrows == 1:
+        axes_array = axes_array.reshape(1, -1)
+    elif ncols == 1:
+        axes_array = axes_array.reshape(-1, 1)
+
+    # Apply background to figure
+    if transparent:
+        fig.patch.set_facecolor("none")
+    elif o.get("facecolor"):
+        fig.patch.set_facecolor(o["facecolor"])
+
+    # Build mapping from axis ID to row/col
+    ax_to_rowcol = _build_ax_to_rowcol_map(axes_spec)
+
+    # Map axes by their ID
+    axes_map = {}
+    for ax_id, ax_spec in axes_spec.items():
+        row, col = ax_to_rowcol.get(ax_id, (0, 0))
+        ax = axes_array[row, col]
+        axes_map[ax_id] = ax
+
+        # Apply background
+        if transparent:
+            ax.patch.set_facecolor("none")
+        elif o.get("facecolor"):
+            ax.patch.set_facecolor(o["facecolor"])
+
+        # Plot data - check which schema format
+        if csv_data is not None:
+            calls = ax_spec.get("calls", [])
+            if calls:
+                # Recipe schema with explicit calls
+                plot_from_recipe(
+                    ax, csv_data, ax_spec, overrides, linewidth_pt, ax_id=ax_id
+                )
+            else:
+                # Editable schema - plot from CSV column names
+                csv_row, csv_col = row, col  # Use computed row/col for CSV lookup
+                _plot_from_editable_csv(
+                    ax,
+                    csv_data,
+                    ax_id,
+                    csv_row,
+                    csv_col,
+                    overrides,
+                    linewidth_pt,
+                    metadata,
+                )
+
+        # Get panel-specific overrides (e.g., ax_00_panel)
+        panel_key = f"{ax_id}_panel"
+        element_overrides = o.get("element_overrides", {})
+        panel_overrides = element_overrides.get(panel_key, {})
+
+        # Apply axis labels from spec, with panel overrides taking precedence
+        xaxis = ax_spec.get("xaxis", {})
+        yaxis = ax_spec.get("yaxis", {})
+
+        # Panel title (from overrides or spec)
+        panel_title = panel_overrides.get("title")
+        if panel_title:
+            ax.set_title(panel_title, fontsize=title_fontsize)
+
+        # X/Y labels (panel overrides take precedence over spec)
+        xlabel = panel_overrides.get("xlabel") or xaxis.get("label")
+        ylabel = panel_overrides.get("ylabel") or yaxis.get("label")
+
+        if xlabel:
+            ax.set_xlabel(xlabel, fontsize=axis_fontsize)
+        if ylabel:
+            ax.set_ylabel(ylabel, fontsize=axis_fontsize)
+
+        # Apply axis limits
+        if xaxis.get("lim"):
+            ax.set_xlim(xaxis["lim"])
+        if yaxis.get("lim"):
+            ax.set_ylim(yaxis["lim"])
+
+        # Tick styling
+        _apply_tick_styling(
+            ax,
+            tick_fontsize,
+            tick_length_pt,
+            tick_width_pt,
+            tick_direction,
+            x_n_ticks,
+            y_n_ticks,
+            False,  # hide_x_ticks
+            False,  # hide_y_ticks
+        )
+
+        # Apply spines
+        if o.get("hide_top_spine", True):
+            ax.spines["top"].set_visible(False)
+        if o.get("hide_right_spine", True):
+            ax.spines["right"].set_visible(False)
+        for spine in ax.spines.values():
+            spine.set_linewidth(axis_width_pt)
+
+        # Apply dark mode to this axis
+        if dark_mode:
+            _apply_dark_theme(ax)
+
+    # Apply caption (below figure) - use global overrides
+    caption_artist = _apply_caption(fig, o)
+    if dark_mode and caption_artist:
+        caption_artist.set_color(DARK_THEME_TEXT_COLOR)
+
+    fig.tight_layout()
+
+    # Get element bounding boxes
+    fig.canvas.draw()
+    renderer = fig.canvas.get_renderer()
+
+    # Save to buffer
+    buf = io.BytesIO()
+    fig.savefig(
+        buf, format="png", dpi=dpi, bbox_inches="tight", transparent=transparent
+    )
+    buf.seek(0)
+
+    # Get actual saved image dimensions
+    img = Image.open(buf)
+    img_width, img_height = img.size
+    buf.seek(0)
+
+    # Get bboxes for all axes
+    bboxes = extract_bboxes_multi(fig, axes_map, renderer, img_width, img_height)
+
+    img_data = base64.b64encode(buf.read()).decode("utf-8")
+    plt.close(fig)
+
+    return img_data, bboxes, {"width": img_width, "height": img_height}
+
+
+def _get_grid_dimensions(axes_spec: Dict[str, Any]) -> Tuple[int, int]:
+    """Get grid dimensions from axes specifications.
+
+    Handles both:
+    - New recipe schema with grid_position
+    - Editable schema - infer grid from positions (y-position groups = rows)
+    """
+    # Check if any axis has grid_position
+    has_grid_pos = any(ax_spec.get("grid_position") for ax_spec in axes_spec.values())
+
+    if has_grid_pos:
+        max_row = 0
+        max_col = 0
+        for ax_id, ax_spec in axes_spec.items():
+            pos = ax_spec.get("grid_position", {})
+            max_row = max(max_row, pos.get("row", 0))
+            max_col = max(max_col, pos.get("col", 0))
+        return max_row + 1, max_col + 1
+
+    # Editable schema: infer grid from positions
+    # Group by y-position to determine rows
+    if not axes_spec:
+        return 1, 1
+
+    positions = []
+    for ax_id, ax_spec in axes_spec.items():
+        pos = ax_spec.get("position", [0, 0, 0, 0])
+        if len(pos) >= 2:
+            positions.append((ax_id, pos[0], pos[1]))  # (id, x, y)
+
+    if not positions:
+        return 1, 1
+
+    # Cluster by y-position (tolerance for floating point)
+    y_values = sorted(set(round(p[2], 2) for p in positions), reverse=True)
+    n_rows = len(y_values)
+
+    # Count columns in first row
+    first_row_y = y_values[0]
+    n_cols = sum(1 for p in positions if abs(p[2] - first_row_y) < 0.1)
+
+    return n_rows, n_cols
+
+
+def _build_ax_to_rowcol_map(axes_spec: Dict[str, Any]) -> Dict[str, Tuple[int, int]]:
+    """Build mapping from axis ID to (row, col) based on positions."""
+    if not axes_spec:
+        return {}
+
+    # Check if any axis has grid_position
+    has_grid_pos = any(ax_spec.get("grid_position") for ax_spec in axes_spec.values())
+
+    if has_grid_pos:
+        result = {}
+        for ax_id, ax_spec in axes_spec.items():
+            pos = ax_spec.get("grid_position", {})
+            result[ax_id] = (pos.get("row", 0), pos.get("col", 0))
+        return result
+
+    # Editable schema: compute from positions
+    positions = []
+    for ax_id, ax_spec in axes_spec.items():
+        pos = ax_spec.get("position", [0, 0, 0, 0])
+        if len(pos) >= 2:
+            positions.append((ax_id, pos[0], pos[1]))  # (id, x, y)
+
+    if not positions:
+        return dict.fromkeys(axes_spec, (0, 0))
+
+    # Get unique y-values (rows) - higher y = lower row number (top of figure)
+    y_values = sorted(set(round(p[2], 2) for p in positions), reverse=True)
+    y_to_row = {y: i for i, y in enumerate(y_values)}
+
+    # For each row, sort by x to get column
+    result = {}
+    for row_idx, row_y in enumerate(y_values):
+        row_axes = [(ax_id, x) for ax_id, x, y in positions if abs(y - row_y) < 0.1]
+        row_axes.sort(key=lambda t: t[1])  # Sort by x
+        for col_idx, (ax_id, _) in enumerate(row_axes):
+            result[ax_id] = (row_idx, col_idx)
+
+    return result
+
+
+def _get_row_col_from_ax_id(
+    ax_id: str, ax_map: Optional[Dict[str, Tuple[int, int]]] = None
+) -> Tuple[int, int]:
+    """Extract row and col from axis ID using the mapping."""
+    if ax_map and ax_id in ax_map:
+        return ax_map[ax_id]
+    # Fallback: try parsing ax_XY format
+    import re
+
+    match = re.match(r"ax_(\d)(\d)", ax_id)
+    if match:
+        return int(match.group(1)), int(match.group(2))
+    return 0, 0
+
+
+def _plot_from_editable_csv(
+    ax,
+    csv_data,
+    ax_id: str,
+    row: int,
+    col: int,
+    overrides: Dict[str, Any],
+    linewidth: float,
+    metadata: Optional[Dict[str, Any]] = None,
+):
+    """Plot data from editable schema CSV format.
+
+    CSV columns follow pattern: ax-row-X-col-Y_trace-id-NAME_variable-VAR
+    """
+    import re
+
+    df = csv_data
+    elements = metadata.get("elements", {}) if metadata else {}
+
+    # Find columns for this axis (by row/col)
+    pattern = f"ax-row-{row}-col-{col}_"
+    ax_cols = [c for c in df.columns if c.startswith(pattern)]
+
+    if not ax_cols:
+        return
+
+    # Group columns by trace-id
+    traces = {}
+    for col_name in ax_cols:
+        # Parse: ax-row-X-col-Y_trace-id-NAME_variable-VAR
+        match = re.match(rf"{pattern}trace-id-(.+?)_variable-(.+)", col_name)
+        if match:
+            trace_id = match.group(1)
+            var_name = match.group(2)
+            if trace_id not in traces:
+                traces[trace_id] = {}
+            traces[trace_id][var_name] = col_name
+
+    # Get element overrides from overrides dict
+    element_overrides = overrides.get("element_overrides", {})
+
+    # Plot each trace
+    trace_idx = 0
+    for trace_id, vars_dict in traces.items():
+        # Find element info from metadata
+        element_key = f"{ax_id}_line_{trace_idx:02d}"
+        element_info = elements.get(element_key, {})
+        label = element_info.get("label", trace_id)
+
+        # Get user overrides for this trace
+        override_key = f"{ax_id}_trace_{trace_idx}"
+        trace_overrides = element_overrides.get(override_key, {})
+
+        # Determine plot type based on variables present
+        # Check more specific patterns first, then fall back to x/y
+
+        if (
+            "y_lower" in vars_dict
+            and "y_middle" in vars_dict
+            and "y_upper" in vars_dict
+        ):
+            # Shaded line plot (fill_between + line)
+            x_col = vars_dict.get("x")
+            if x_col:
+                x = df[x_col].dropna().values
+                y_lower = df[vars_dict["y_lower"]].dropna().values
+                y_middle = df[vars_dict["y_middle"]].dropna().values
+                y_upper = df[vars_dict["y_upper"]].dropna().values
+
+                min_len = min(len(x), len(y_lower), len(y_middle), len(y_upper))
+                if min_len > 0:
+                    ax.fill_between(
+                        x[:min_len], y_lower[:min_len], y_upper[:min_len], alpha=0.3
+                    )
+                    ax.plot(x[:min_len], y_middle[:min_len], linewidth=linewidth)
+            trace_idx += 1
+
+        elif "row" in vars_dict and "col" in vars_dict and "value" in vars_dict:
+            # Heatmap / imshow
+            rows_data = df[vars_dict["row"]].dropna().values
+            cols_data = df[vars_dict["col"]].dropna().values
+            values = df[vars_dict["value"]].dropna().values
+            if len(rows_data) > 0:
+                n_rows = int(rows_data.max()) + 1
+                n_cols = int(cols_data.max()) + 1
+                data = np.zeros((n_rows, n_cols))
+                for r, c, v in zip(
+                    rows_data.astype(int), cols_data.astype(int), values
+                ):
+                    data[r, c] = v
+                ax.imshow(data, aspect="auto", origin="lower")
+
+        elif "y1" in vars_dict and "y2" in vars_dict:
+            # fill_between (CI band)
+            x_col = vars_dict.get("x")
+            if x_col:
+                x = df[x_col].dropna().values
+                y1 = df[vars_dict["y1"]].dropna().values
+                y2 = df[vars_dict["y2"]].dropna().values
+                min_len = min(len(x), len(y1), len(y2))
+                if min_len > 0:
+                    ax.fill_between(x[:min_len], y1[:min_len], y2[:min_len], alpha=0.3)
+
+        elif "yerr" in vars_dict and "y" in vars_dict:
+            # Error bars with bar chart
+            x_col = vars_dict.get("x")
+            y_col = vars_dict.get("y")
+            if x_col and y_col:
+                x = df[x_col].dropna().values
+                y = df[y_col].dropna().values
+                yerr = df[vars_dict["yerr"]].dropna().values
+                min_len = min(len(x), len(y), len(yerr))
+                if min_len > 0:
+                    ax.bar(x[:min_len], y[:min_len], yerr=yerr[:min_len])
+
+        elif "group" in vars_dict and "value" in vars_dict:
+            # Violin/strip plot - plot as scatter for now
+            groups = df[vars_dict["group"]].dropna().values
+            values = df[vars_dict["value"]].dropna().values
+            if len(groups) > 0 and len(values) > 0:
+                min_len = min(len(groups), len(values))
+                # Convert string groups to numeric positions
+                unique_groups = list(dict.fromkeys(groups[:min_len]))  # Preserve order
+                group_to_x = {g: i for i, g in enumerate(unique_groups)}
+                x_positions = np.array([group_to_x.get(g, 0) for g in groups[:min_len]])
+                # Add jitter for strip plot effect
+                jitter = np.random.uniform(-0.1, 0.1, min_len)
+                ax.scatter(x_positions + jitter, values[:min_len], alpha=0.6, s=20)
+                # Set tick labels
+                ax.set_xticks(range(len(unique_groups)))
+                ax.set_xticklabels(unique_groups, fontsize=6)
+
+        elif "width" in vars_dict and "height" in vars_dict:
+            # Rectangle - skip for now
+            pass
+
+        elif "type" in vars_dict:
+            # Skip type-only entries (like stim markers)
+            pass
+
+        elif "content" in vars_dict:
+            # Text annotation - skip for preview
+            pass
+
+        elif "x" in vars_dict and "y" in vars_dict:
+            # Default: line or scatter plot
+            x_col = vars_dict["x"]
+            y_col = vars_dict["y"]
+            x = df[x_col].dropna().values
+            y = df[y_col].dropna().values
+            if len(x) > 0 and len(y) > 0:
+                min_len = min(len(x), len(y))
+                # Apply overrides
+                color = trace_overrides.get("color")
+                lw = trace_overrides.get("linewidth", linewidth)
+                ls = trace_overrides.get("linestyle", "-")
+                marker = trace_overrides.get("marker")
+                alpha = trace_overrides.get("alpha", 1.0)
+
+                kwargs = {"linewidth": lw, "linestyle": ls, "alpha": alpha}
+                if color:
+                    kwargs["color"] = color
+                if marker:
+                    kwargs["marker"] = marker
+                if label and label != trace_id:
+                    kwargs["label"] = label
+
+                # Check if this looks like scatter data (trace-id contains 'scatter' or 'strip')
+                if "scatter" in trace_id.lower() or "strip" in trace_id.lower():
+                    scatter_kwargs = {"alpha": alpha, "s": 20}
+                    if color:
+                        scatter_kwargs["c"] = color
+                    ax.scatter(x[:min_len], y[:min_len], **scatter_kwargs)
+                else:
+                    ax.plot(x[:min_len], y[:min_len], **kwargs)
+            trace_idx += 1
+
+
+def _apply_background(fig, ax, o, transparent):
+    """Apply background settings to figure."""
+    if transparent:
+        fig.patch.set_facecolor("none")
+        ax.patch.set_facecolor("none")
+    elif o.get("facecolor"):
+        fig.patch.set_facecolor(o["facecolor"])
+        ax.patch.set_facecolor(o["facecolor"])
+
+
+def _apply_labels(ax, o, title_fontsize, axis_fontsize):
+    """Apply title and axis labels."""
+    # Show title only if enabled (default True)
+    if o.get("show_title", True) and o.get("title"):
+        ax.set_title(o["title"], fontsize=title_fontsize)
+    if o.get("xlabel"):
+        ax.set_xlabel(o["xlabel"], fontsize=axis_fontsize)
+    if o.get("ylabel"):
+        ax.set_ylabel(o["ylabel"], fontsize=axis_fontsize)
+
+
+def _apply_caption(fig, o, caption_fontsize=7):
+    """Apply caption below the figure."""
+    if not o.get("show_caption", False) or not o.get("caption"):
+        return None
+
+    caption_text = o.get("caption", "")
+    fontsize = o.get("caption_fontsize", caption_fontsize)
+
+    # Place caption below the figure
+    # Using fig.text with y position slightly below 0
+    caption_artist = fig.text(
+        0.5,
+        -0.02,  # Centered, below the figure
+        caption_text,
+        ha="center",
+        va="top",
+        fontsize=fontsize,
+        wrap=True,
+        transform=fig.transFigure,
+    )
+    return caption_artist
+
+
+def _apply_tick_styling(
+    ax,
+    tick_fontsize,
+    tick_length_pt,
+    tick_width_pt,
+    tick_direction,
+    x_n_ticks,
+    y_n_ticks,
+    hide_x_ticks,
+    hide_y_ticks,
+):
+    """Apply tick styling to axes."""
+    ax.tick_params(
+        axis="both",
+        labelsize=tick_fontsize,
+        length=tick_length_pt,
+        width=tick_width_pt,
+        direction=tick_direction,
+    )
+
+    if hide_x_ticks:
+        ax.xaxis.set_ticks([])
+        ax.xaxis.set_ticklabels([])
+    else:
+        ax.xaxis.set_major_locator(MaxNLocator(nbins=x_n_ticks))
+    if hide_y_ticks:
+        ax.yaxis.set_ticks([])
+        ax.yaxis.set_ticklabels([])
+    else:
+        ax.yaxis.set_major_locator(MaxNLocator(nbins=y_n_ticks))
+
+
+def _apply_style(ax, o, axis_width_pt):
+    """Apply grid, axis limits, and spine settings."""
+    if o.get("grid"):
+        ax.grid(True, linewidth=axis_width_pt, alpha=0.3)
+
+    if o.get("xlim"):
+        ax.set_xlim(o["xlim"])
+    if o.get("ylim"):
+        ax.set_ylim(o["ylim"])
+
+    if o.get("hide_top_spine", True):
+        ax.spines["top"].set_visible(False)
+    if o.get("hide_right_spine", True):
+        ax.spines["right"].set_visible(False)
+
+    for spine in ax.spines.values():
+        spine.set_linewidth(axis_width_pt)
+
+
+def _apply_annotations(ax, o, axis_fontsize):
+    """Apply text annotations to figure."""
+    for annot in o.get("annotations", []):
+        if annot.get("type") == "text":
+            ax.text(
+                annot.get("x", 0.5),
+                annot.get("y", 0.5),
+                annot.get("text", ""),
+                transform=ax.transAxes,
+                fontsize=annot.get("fontsize", axis_fontsize),
+            )
+
+
+def render_panel_preview(
+    panel_dir,
+    dark_mode: bool = False,
+) -> Tuple[Optional[str], Optional[Dict[str, Any]], Optional[Dict[str, int]]]:
+    """Render a panel from its pltz bundle directory with dark mode support.
+
+    Args:
+        panel_dir: Path to the .pltz.d panel directory
+        dark_mode: Whether to render with dark mode colors
+
+    Returns:
+        tuple: (base64_image_data, bboxes_dict, image_size) or (None, None, None) on error
+    """
+    import json
+    from pathlib import Path
+
+    import pandas as pd
+
+    panel_dir = Path(panel_dir)
+
+    try:
+        # Load spec.json
+        spec_path = panel_dir / "spec.json"
+        if not spec_path.exists():
+            # Try legacy format
+            for f in panel_dir.glob("*.json"):
+                if f.name != "style.json":
+                    spec_path = f
+                    break
+
+        if not spec_path.exists():
+            return None, None, None
+
+        with open(spec_path) as f:
+            metadata = json.load(f)
+
+        # Load CSV data
+        csv_data = None
+        csv_path = panel_dir / "data.csv"
+        if not csv_path.exists():
+            for f in panel_dir.glob("*.csv"):
+                csv_path = f
+                break
+
+        if csv_path.exists():
+            csv_data = pd.read_csv(csv_path)
+
+        # Load style.json for overrides
+        style_path = panel_dir / "style.json"
+        overrides = {}
+        if style_path.exists():
+            with open(style_path) as f:
+                style = json.load(f)
+            # Convert style to overrides format
+            size = style.get("size", {})
+            if size:
+                width_mm = size.get("width_mm", 80)
+                height_mm = size.get("height_mm", 68)
+                overrides["fig_size"] = [width_mm / 25.4, height_mm / 25.4]
+            overrides["transparent"] = True
+
+        # Render with dark mode
+        return render_preview_with_bboxes(
+            csv_data,
+            overrides,
+            metadata=metadata,
+            dark_mode=dark_mode,
+        )
+
+    except Exception as e:
+        import traceback
+
+        print(f"Error rendering panel {panel_dir}: {e}")
+        traceback.print_exc()
+        return None, None, None
+
+
+# Dark mode theme colors
+DARK_THEME_TEXT_COLOR = "#e8e8e8"  # Light gray for visibility on dark background
+DARK_THEME_SPINE_COLOR = "#e8e8e8"
+DARK_THEME_TICK_COLOR = "#e8e8e8"
+
+
+def _apply_dark_theme(ax):
+    """Apply dark mode colors to axes for visibility on dark backgrounds.
+
+    Changes title, labels, tick labels, spines, and legend text to light colors.
+    """
+    # Title
+    title = ax.get_title()
+    if title:
+        ax.title.set_color(DARK_THEME_TEXT_COLOR)
+
+    # Axis labels
+    ax.xaxis.label.set_color(DARK_THEME_TEXT_COLOR)
+    ax.yaxis.label.set_color(DARK_THEME_TEXT_COLOR)
+
+    # Tick labels
+    ax.tick_params(
+        axis="both", colors=DARK_THEME_TICK_COLOR, labelcolor=DARK_THEME_TEXT_COLOR
+    )
+
+    # Spines
+    for spine in ax.spines.values():
+        spine.set_color(DARK_THEME_SPINE_COLOR)
+
+    # Legend (if exists)
+    legend = ax.get_legend()
+    if legend:
+        for text in legend.get_texts():
+            text.set_color(DARK_THEME_TEXT_COLOR)
+        # Legend title
+        legend_title = legend.get_title()
+        if legend_title:
+            legend_title.set_color(DARK_THEME_TEXT_COLOR)
+        # Legend frame (make transparent or dark)
+        legend.get_frame().set_facecolor("none")
+        legend.get_frame().set_edgecolor(DARK_THEME_SPINE_COLOR)
+
+
+# EOF