opik 1.6.4__py3-none-any.whl → 1.9.71__py3-none-any.whl

opik/cli/usage_report/charts.py

@@ -0,0 +1,783 @@
+"""Chart creation functions for usage report module."""
+
+import datetime
+import os
+import time
+import traceback
+from typing import Any, Dict, List, Optional, Tuple
+
+from rich.console import Console
+
+from .utils import extract_metric_data
+
+console = Console()
+
+
+def _get_top_projects_and_others(
+    projects: List[Dict[str, Any]],
+    project_names: List[str],
+    metric_data: List[List[float]],
+    top_n: int = 12,
+) -> Tuple[List[int], List[float], List[str], List]:
+    """
+    Identify top N projects by total usage and group the rest as "Others".
+
+    Args:
+        projects: List of project dictionaries
+        project_names: List of project names
+        metric_data: List of lists, where each inner list contains values for one period
+        top_n: Number of top projects to show individually
+
+    Returns:
+        Tuple of (top_project_indices, others_data, labels, colors)
+        - top_project_indices: List of indices for top projects
+        - others_data: List of aggregated values for "Others" per period
+        - labels: List of labels (top project names + "Others")
+        - colors: List of colors for top projects + "Others"
+    """
+    try:
+        import matplotlib.colors as mcolors
+        import matplotlib.pyplot as plt
+    except ImportError:
+        raise ImportError(
+            "matplotlib is required for chart generation. "
+            "Please install it with: pip install matplotlib"
+        )
+
+    # Calculate total usage per project across all periods
+    project_totals = []
+    for i in range(len(project_names)):
+        total = sum(metric_data[j][i] for j in range(len(metric_data)))
+        project_totals.append((i, total))
+
+    # Sort by total (descending) and get top N
+    project_totals.sort(key=lambda x: x[1], reverse=True)
+    top_indices = [idx for idx, _ in project_totals[:top_n]]
+    others_indices = [idx for idx, _ in project_totals[top_n:]]
+
+    # Create labels
+    labels = [project_names[i] for i in top_indices]
+    if others_indices:
+        labels.append(f"Others ({len(others_indices)} projects)")
+
+    # Aggregate "Others" data
+    others_data = []
+    if others_indices:
+        for period_idx in range(len(metric_data)):
+            others_total = sum(metric_data[period_idx][i] for i in others_indices)
+            others_data.append(others_total)
+    else:
+        others_data = [0.0] * len(metric_data)
+
+    # Generate colors for top projects + Others
+    colors_list = []
+    colormaps = [
+        plt.cm.tab20,
+        plt.cm.tab20b,
+        plt.cm.Set3,
+        plt.cm.Pastel1,
+        plt.cm.Pastel2,
+        plt.cm.Set1,
+        plt.cm.Set2,
+    ]
+
+    for i in range(len(top_indices)):
+        if i < 20:
+            colors_list.append(colormaps[0](i))
+        elif i < 40:
+            colors_list.append(colormaps[1](i - 20))
+        elif i < 52:
+            colors_list.append(colormaps[2]((i - 40) % 12))
+        elif i < 61:
+            colors_list.append(colormaps[3]((i - 52) % 9))
+        elif i < 69:
+            colors_list.append(colormaps[4]((i - 61) % 8))
+        elif i < 78:
+            colors_list.append(colormaps[5]((i - 69) % 9))
+        elif i < 86:
+            colors_list.append(colormaps[6]((i - 78) % 8))
+        else:
+            hue = (i * 0.618033988749895) % 1.0
+            saturation = 0.6 + (i % 3) * 0.1
+            value = 0.85 + (i % 2) * 0.1
+            colors_list.append(mcolors.hsv_to_rgb([hue, saturation, value]))
+
+    # Add gray color for "Others"
+    if others_indices:
+        colors_list.append("#808080")  # Gray for Others
+
+    return top_indices, others_data, labels, colors_list
+
+
+def create_charts(data: Dict[str, Any], output_dir: str = ".") -> None:
+    """
+    Create stacked bar charts for trace count, token count, cost, experiment count, and dataset count.
+
+    Note: This function creates charts in memory but does not save them to disk.
+    Charts are generated and immediately closed. For saving charts, use create_individual_chart()
+    which is used by the PDF report generation.
+
+    Args:
+        data: The extracted data dictionary
+        output_dir: Directory parameter (kept for backward compatibility, not used)
+    """
+    try:
+        import matplotlib.pyplot as plt
+        from matplotlib.ticker import FuncFormatter
+    except ImportError:
+        raise ImportError(
+            "matplotlib is required for chart generation. "
+            "Please install it with: pip install matplotlib"
+        )
+
+    # Get unit from data (default to month for backward compatibility)
+    unit = data.get("unit", "month")
+
+    # Prepare data for charts
+    projects = [
+        p for p in data["projects"] if "metrics_by_unit" in p and "error" not in p
+    ]
+    if not projects:
+        console.print("[yellow]No project data available for charting.[/yellow]")
+        return
+
+    # Collect all time periods across all projects
+    all_periods_set = set()
+    for project in projects:
+        all_periods_set.update(project["metrics_by_unit"].keys())
+    all_periods: List[str] = sorted(all_periods_set)
+
+    if not all_periods:
+        console.print(f"[yellow]No {unit}ly data available for charting.[/yellow]")
+        return
+
+    # Prepare data arrays for each metric
+    project_names = [p["project_name"] for p in projects]
+    n_periods = len(all_periods)
+
+    # Helper function for token count aggregation
+    def aggregate_token_count(token_count: Any) -> float:
+        """Aggregate token count: use total_tokens if available, otherwise sum all values."""
+        if isinstance(token_count, dict):
+            if "total_tokens" in token_count:
+                return float(token_count["total_tokens"])
+            else:
+                return (
+                    sum(float(v) for v in token_count.values()) if token_count else 0.0
+                )
+        else:
+            return float(token_count) if token_count else 0.0
+
+    # Extract metric data using helper function
+    trace_data = extract_metric_data(projects, all_periods, "trace_count")
+    token_data = extract_metric_data(
+        projects, all_periods, "token_count", aggregate_token_count
+    )
+    cost_data = extract_metric_data(projects, all_periods, "cost")
+    span_data = extract_metric_data(projects, all_periods, "span_count")
+
+    # Format period labels for display based on unit
+    period_labels = []
+    for period in all_periods:
+        if unit == "month":
+            period_labels.append(
+                datetime.datetime.strptime(period, "%Y-%m").strftime("%b %Y")
+            )
+        elif unit == "week":
+            # Parse ISO week format: YYYY-Www
+            try:
+                if "-W" in period:
+                    year, week = period.split("-W", 1)
+                    period_labels.append(f"Week {week}, {year}")
+                else:
+                    period_labels.append(period)
+            except (ValueError, IndexError):
+                period_labels.append(period)
+        elif unit == "day":
+            period_labels.append(
+                datetime.datetime.strptime(period, "%Y-%m-%d").strftime("%b %d, %Y")
+            )
+        elif unit == "hour":
+            period_labels.append(
+                datetime.datetime.strptime(period, "%Y-%m-%d-%H").strftime(
+                    "%b %d, %Y %H:00"
+                )
+            )
+        else:
+            period_labels.append(period)
+
+    # Get experiment data (workspace-level)
+    experiment_data = []
+    for period in all_periods:
+        experiment_count = data.get("experiments_by_unit", {}).get(period, 0)
+        experiment_data.append(float(experiment_count) if experiment_count else 0.0)
+
+    # Get dataset data (workspace-level)
+    dataset_data = []
+    for period in all_periods:
+        dataset_count = data.get("datasets_by_unit", {}).get(period, 0)
+        dataset_data.append(float(dataset_count) if dataset_count else 0.0)
+
+    # Create figure with 6 subplots
+    # Increase height to give more room for charts (less space needed for legend now)
+    fig, axes = plt.subplots(6, 1, figsize=(14, 20))
+    unit_label = unit.capitalize()
+    fig.suptitle(
+        f'Opik Usage Metrics - {data["workspace"]} (by {unit_label})',
+        fontsize=16,
+        fontweight="bold",
+    )
+
+    # Chart 1: Trace Count - use top projects only
+    ax1 = axes[0]
+    x = range(n_periods)
+    width = 0.8
+
+    # Get top projects for trace count
+    top_indices, others_data, trace_labels, trace_colors = _get_top_projects_and_others(
+        projects, project_names, trace_data, top_n=18
+    )
+
+    bottom = [0] * n_periods
+    for idx, (project_idx, label) in enumerate(
+        zip(top_indices, trace_labels[: len(top_indices)])
+    ):
+        values: List[float] = [trace_data[j][project_idx] for j in range(n_periods)]
+        ax1.bar(x, values, width, label=label, bottom=bottom, color=trace_colors[idx])
+        bottom = [float(bottom[j] + values[j]) for j in range(n_periods)]  # type: ignore[misc]
+
+    # Add "Others" if present
+    if others_data and any(v > 0 for v in others_data):
+        ax1.bar(
+            x,
+            others_data,
+            width,
+            label=trace_labels[-1],
+            bottom=bottom,
+            color=trace_colors[-1],
+        )
+
+    ax1.set_xlabel(unit_label)
+    ax1.set_ylabel("Trace Count")
+    ax1.set_title(f"Trace Count by {unit_label} (Stacked by Project)")
+    ax1.set_xticks(x)
+    ax1.set_xticklabels(period_labels, rotation=45, ha="right")
+    ax1.legend(
+        bbox_to_anchor=(0.5, -0.20),
+        loc="upper center",
+        ncol=4,
+        fontsize=7,
+        frameon=True,
+    )
+    ax1.grid(axis="y", alpha=0.3)
+
+    # Chart 2: Token Count - use top projects only
+    ax2 = axes[1]
+
+    # Get top projects for token count
+    top_indices, others_data, token_labels, token_colors = _get_top_projects_and_others(
+        projects, project_names, token_data, top_n=18
+    )
+
+    bottom = [0] * n_periods
+    for idx, (project_idx, label) in enumerate(
+        zip(top_indices, token_labels[: len(top_indices)])
+    ):
+        values: List[float] = [token_data[j][project_idx] for j in range(n_periods)]  # type: ignore[no-redef]
+        ax2.bar(x, values, width, label=label, bottom=bottom, color=token_colors[idx])
+        bottom = [float(bottom[j] + values[j]) for j in range(n_periods)]  # type: ignore[misc]
+
+    # Add "Others" if present
+    if others_data and any(v > 0 for v in others_data):
+        ax2.bar(
+            x,
+            others_data,
+            width,
+            label=token_labels[-1],
+            bottom=bottom,
+            color=token_colors[-1],
+        )
+
+    ax2.set_xlabel(unit_label)
+    ax2.set_ylabel("Token Count")
+    ax2.set_title(f"Token Count by {unit_label} (Stacked by Project)")
+    ax2.set_xticks(x)
+    ax2.set_xticklabels(period_labels, rotation=45, ha="right")
+    ax2.legend(
+        bbox_to_anchor=(0.5, -0.20),
+        loc="upper center",
+        ncol=4,
+        fontsize=7,
+        frameon=True,
+    )
+    ax2.grid(axis="y", alpha=0.3)
+    # Format y-axis to show in thousands/millions
+    ax2.yaxis.set_major_formatter(
+        FuncFormatter(
+            lambda x, p: (
+                f"{x/1e6:.2f}M"
+                if x >= 1e6
+                else f"{x/1e3:.0f}K"
+                if x >= 1e3
+                else f"{x:.0f}"
+            )
+        )
+    )
+
+    # Chart 3: Cost - use top projects only
+    ax3 = axes[2]
+
+    # Get top projects for cost
+    top_indices, others_data, cost_labels, cost_colors = _get_top_projects_and_others(
+        projects, project_names, cost_data, top_n=18
+    )
+
+    bottom = [0] * n_periods
+    for idx, (project_idx, label) in enumerate(
+        zip(top_indices, cost_labels[: len(top_indices)])
+    ):
+        values: List[float] = [cost_data[j][project_idx] for j in range(n_periods)]  # type: ignore[no-redef]
+        ax3.bar(x, values, width, label=label, bottom=bottom, color=cost_colors[idx])
+        bottom = [float(bottom[j] + values[j]) for j in range(n_periods)]  # type: ignore[misc]
+
+    # Add "Others" if present
+    if others_data and any(v > 0 for v in others_data):
+        ax3.bar(
+            x,
+            others_data,
+            width,
+            label=cost_labels[-1],
+            bottom=bottom,
+            color=cost_colors[-1],
+        )
+
+    ax3.set_xlabel(unit_label)
+    ax3.set_ylabel("Cost ($)")
+    ax3.set_title(f"Cost by {unit_label} (Stacked by Project)")
+    ax3.set_xticks(x)
+    ax3.set_xticklabels(period_labels, rotation=45, ha="right")
+    ax3.legend(
+        bbox_to_anchor=(0.5, -0.20),
+        loc="upper center",
+        ncol=4,
+        fontsize=7,
+        frameon=True,
+    )
+    ax3.grid(axis="y", alpha=0.3)
+    # Format y-axis for currency
+    ax3.yaxis.set_major_formatter(FuncFormatter(lambda x, p: f"${x:.2f}"))
+
+    # Chart 4: Experiment Count (workspace-level, not stacked)
+    ax4 = axes[3]
+    ax4.bar(x, experiment_data, width, color="steelblue", alpha=0.7)
+
+    ax4.set_xlabel(unit_label)
+    ax4.set_ylabel("Experiment Count")
+    ax4.set_title(f"Experiment Count by {unit_label} (Workspace Total)")
+    ax4.set_xticks(x)
+    ax4.set_xticklabels(period_labels, rotation=45, ha="right")
+    ax4.grid(axis="y", alpha=0.3)
+
+    # Chart 5: Dataset Count (workspace-level, not stacked)
+    ax5 = axes[4]
+    ax5.bar(x, dataset_data, width, color="darkgreen", alpha=0.7)
+
+    ax5.set_xlabel(unit_label)
+    ax5.set_ylabel("Dataset Count")
+    ax5.set_title(f"Dataset Count by {unit_label} (Workspace Total)")
+    ax5.set_xticks(x)
+    ax5.set_xticklabels(period_labels, rotation=45, ha="right")
+    ax5.grid(axis="y", alpha=0.3)
+
+    # Chart 6: Span Count - use top projects only
+    ax6 = axes[5]
+
+    # Get top projects for span count
+    top_indices, others_data, span_labels, span_colors = _get_top_projects_and_others(
+        projects, project_names, span_data, top_n=18
+    )
+
+    bottom = [0] * n_periods
+    for idx, (project_idx, label) in enumerate(
+        zip(top_indices, span_labels[: len(top_indices)])
+    ):
+        values: List[float] = [span_data[j][project_idx] for j in range(n_periods)]  # type: ignore[no-redef]
+        ax6.bar(x, values, width, label=label, bottom=bottom, color=span_colors[idx])
+        bottom = [float(bottom[j] + values[j]) for j in range(n_periods)]  # type: ignore[misc]
+
+    # Add "Others" if present
+    if others_data and any(v > 0 for v in others_data):
+        ax6.bar(
+            x,
+            others_data,
+            width,
+            label=span_labels[-1],
+            bottom=bottom,
+            color=span_colors[-1],
+        )
+
+    ax6.set_xlabel(unit_label)
+    ax6.set_ylabel("Span Count")
+    ax6.set_title(f"Span Count by {unit_label} (Stacked by Project)")
+    ax6.set_xticks(x)
+    ax6.set_xticklabels(period_labels, rotation=45, ha="right")
+    ax6.legend(
+        bbox_to_anchor=(0.5, -0.20),
+        loc="upper center",
+        ncol=4,
+        fontsize=7,
+        frameon=True,
+    )
+    ax6.grid(axis="y", alpha=0.3)
+
+    # Use rect parameter to make room for legends below charts (more space for lower legends)
+    plt.tight_layout(rect=[0, 0.0, 1, 0.98])
+
+    plt.close()
+
+
+def create_individual_chart(
+    data: Dict[str, Any],
+    chart_type: str,
+    output_dir: str = ".",
+) -> Optional[str]:
+    """
+    Create an individual chart figure for a specific chart type.
+
+    Args:
+        data: The extracted data dictionary
+        chart_type: Type of chart - "trace_count", "token_count", "cost", "experiment_count", "dataset_count", "span_count"
+        output_dir: Directory to save chart (default: current directory)
+
+    Returns:
+        Path to saved chart image file, or None if creation failed
+    """
+    try:
+        import matplotlib.pyplot as plt
+        from matplotlib.ticker import FuncFormatter
+    except ImportError:
+        raise ImportError(
+            "matplotlib is required for chart generation. "
+            "Please install it with: pip install matplotlib"
+        )
+
+    # Get unit from data (default to month for backward compatibility)
+    unit = data.get("unit", "month")
+
+    # Prepare data for charts
+    projects = [
+        p for p in data["projects"] if "metrics_by_unit" in p and "error" not in p
+    ]
+    if not projects:
+        return None
+
+    # Collect all time periods across all projects
+    all_periods_set = set()
+    for project in projects:
+        all_periods_set.update(project["metrics_by_unit"].keys())
+    all_periods: List[str] = sorted(all_periods_set)
+
+    if not all_periods:
+        return None
+
+    # Prepare data arrays for each metric
+    project_names = [p["project_name"] for p in projects]
+    n_periods = len(all_periods)
+
+    # Format period labels for display based on unit
+    period_labels = []
+    for period in all_periods:
+        if unit == "month":
+            period_labels.append(
+                datetime.datetime.strptime(period, "%Y-%m").strftime("%b %Y")
+            )
+        elif unit == "week":
+            try:
+                if "-W" in period:
+                    year, week = period.split("-W", 1)
+                    period_labels.append(f"Week {week}, {year}")
+                else:
+                    period_labels.append(period)
+            except (ValueError, IndexError):
+                period_labels.append(period)
+        elif unit == "day":
+            period_labels.append(
+                datetime.datetime.strptime(period, "%Y-%m-%d").strftime("%b %d, %Y")
+            )
+        elif unit == "hour":
+            period_labels.append(
+                datetime.datetime.strptime(period, "%Y-%m-%d-%H").strftime(
+                    "%b %d, %Y %H:00"
+                )
+            )
+        else:
+            period_labels.append(period)
+
+    # Create figure with consistent size for all charts (same as reference implementation)
+    fig, ax = plt.subplots(figsize=(14, 8))
+    unit_label = unit.capitalize()
+    x = range(n_periods)
+    width = 0.8
+
+    if chart_type == "trace_count":
+        # Trace count data
+        trace_data = extract_metric_data(projects, all_periods, "trace_count")
+
+        # Get top projects for trace count
+        top_indices, others_data, labels, colors = _get_top_projects_and_others(
+            projects, project_names, trace_data, top_n=18
+        )
+
+        bottom = [0] * n_periods
+        for idx, (project_idx, label) in enumerate(
+            zip(top_indices, labels[: len(top_indices)])
+        ):
+            values: List[float] = [trace_data[j][project_idx] for j in range(n_periods)]
+            ax.bar(x, values, width, label=label, bottom=bottom, color=colors[idx])
+            bottom = [float(bottom[j] + values[j]) for j in range(n_periods)]  # type: ignore[misc]
+
+        # Add "Others" if present
+        if others_data and any(v > 0 for v in others_data):
+            ax.bar(
+                x, others_data, width, label=labels[-1], bottom=bottom, color=colors[-1]
+            )
+
+        ax.set_ylabel("Trace Count")
+        ax.set_title(f"Trace Count by {unit_label} (Stacked by Project)")
+
+    elif chart_type == "token_count":
+        # Helper function for token count aggregation
+        def aggregate_token_count(token_count: Any) -> float:
+            """Aggregate token count: use total_tokens if available, otherwise sum all values."""
+            if isinstance(token_count, dict):
+                if "total_tokens" in token_count:
+                    return float(token_count["total_tokens"])
+                else:
+                    return (
+                        sum(float(v) for v in token_count.values())
+                        if token_count
+                        else 0.0
+                    )
+            else:
+                return float(token_count) if token_count else 0.0
+
+        # Token count data
+        token_data = extract_metric_data(
+            projects, all_periods, "token_count", aggregate_token_count
+        )
+
+        # Get top projects for token count
+        top_indices, others_data, labels, colors = _get_top_projects_and_others(
+            projects, project_names, token_data, top_n=18
+        )
+
+        bottom = [0] * n_periods
+        for idx, (project_idx, label) in enumerate(
+            zip(top_indices, labels[: len(top_indices)])
+        ):
+            values: List[float] = [token_data[j][project_idx] for j in range(n_periods)]  # type: ignore[no-redef]
+            ax.bar(x, values, width, label=label, bottom=bottom, color=colors[idx])
+            bottom = [float(bottom[j] + values[j]) for j in range(n_periods)]  # type: ignore[misc]
+
+        # Add "Others" if present
+        if others_data and any(v > 0 for v in others_data):
+            ax.bar(
+                x, others_data, width, label=labels[-1], bottom=bottom, color=colors[-1]
+            )
+
+        ax.set_ylabel("Token Count")
+        ax.set_title(f"Token Count by {unit_label} (Stacked by Project)")
+        ax.yaxis.set_major_formatter(
+            FuncFormatter(
+                lambda x, p: (
+                    f"{x/1e6:.2f}M"
+                    if x >= 1e6
+                    else f"{x/1e3:.0f}K"
+                    if x >= 1e3
+                    else f"{x:.0f}"
+                )
+            )
+        )
+
+    elif chart_type == "cost":
+        # Cost data
+        cost_data = extract_metric_data(projects, all_periods, "cost")
+
+        # Get top projects for cost
+        top_indices, others_data, labels, colors = _get_top_projects_and_others(
+            projects, project_names, cost_data, top_n=18
+        )
+
+        bottom = [0] * n_periods
+        for idx, (project_idx, label) in enumerate(
+            zip(top_indices, labels[: len(top_indices)])
+        ):
+            values: List[float] = [cost_data[j][project_idx] for j in range(n_periods)]  # type: ignore[no-redef]
+            ax.bar(x, values, width, label=label, bottom=bottom, color=colors[idx])
+            bottom = [float(bottom[j] + values[j]) for j in range(n_periods)]  # type: ignore[misc]
+
+        # Add "Others" if present
+        if others_data and any(v > 0 for v in others_data):
+            ax.bar(
+                x, others_data, width, label=labels[-1], bottom=bottom, color=colors[-1]
+            )
+
+        ax.set_ylabel("Cost ($)")
+        ax.set_title(f"Cost by {unit_label} (Stacked by Project)")
+        ax.yaxis.set_major_formatter(FuncFormatter(lambda x, p: f"${x:.2f}"))
+
+    elif chart_type == "experiment_count":
+        # Experiment data (workspace-level)
+        experiment_data = []
+        for period in all_periods:
+            experiment_count = data.get("experiments_by_unit", {}).get(period, 0)
+            experiment_data.append(float(experiment_count) if experiment_count else 0.0)
+
+        ax.bar(x, experiment_data, width, color="steelblue", alpha=0.7)
+        ax.set_ylabel("Experiment Count")
+        ax.set_title(f"Experiment Count by {unit_label} (Workspace Total)")
+
+    elif chart_type == "dataset_count":
+        # Dataset data (workspace-level)
+        dataset_data = []
+        for period in all_periods:
+            dataset_count = data.get("datasets_by_unit", {}).get(period, 0)
+            dataset_data.append(float(dataset_count) if dataset_count else 0.0)
+
+        ax.bar(x, dataset_data, width, color="darkgreen", alpha=0.7)
+        ax.set_ylabel("Dataset Count")
+        ax.set_title(f"Dataset Count by {unit_label} (Workspace Total)")
+
+    elif chart_type == "span_count":
+        # Span count data
+        span_data = extract_metric_data(projects, all_periods, "span_count")
+
+        # Get top projects for span count
+        top_indices, others_data, labels, colors = _get_top_projects_and_others(
+            projects, project_names, span_data, top_n=18
+        )
+
+        bottom = [0] * n_periods
+        for idx, (project_idx, label) in enumerate(
+            zip(top_indices, labels[: len(top_indices)])
+        ):
+            values: List[float] = [span_data[j][project_idx] for j in range(n_periods)]  # type: ignore[no-redef]
+            ax.bar(x, values, width, label=label, bottom=bottom, color=colors[idx])
+            bottom = [float(bottom[j] + values[j]) for j in range(n_periods)]  # type: ignore[misc]
+
+        # Add "Others" if present
+        if others_data and any(v > 0 for v in others_data):
+            ax.bar(
+                x, others_data, width, label=labels[-1], bottom=bottom, color=colors[-1]
+            )
+
+        ax.set_ylabel("Span Count")
+        ax.set_title(f"Span Count by {unit_label} (Stacked by Project)")
+
+    else:
+        plt.close()
+        return None
+
+    ax.set_xlabel(unit_label)
+    ax.set_xticks(x)
+    ax.set_xticklabels(period_labels, rotation=45, ha="right")
+    # Set x-axis limits to use full width, with small padding on edges
+    ax.set_xlim(-0.5, n_periods - 0.5)
+
+    ax.grid(axis="y", alpha=0.3)
+
+    # Configure legend for charts that need it - place inside figure bounds
+    has_legend = chart_type in ["trace_count", "token_count", "cost", "span_count"]
+    if has_legend:
+        # Truncate legend labels to maximum length to prevent overly wide legends
+        handles, labels = ax.get_legend_handles_labels()
+        max_label_length = 40  # Maximum characters per legend label
+        truncated_labels = []
+        for label in labels:
+            if len(label) > max_label_length:
+                truncated_labels.append(label[: max_label_length - 3] + "...")
+            else:
+                truncated_labels.append(label)
+
+        # Position legend inside the plot area at the bottom, with more space below
+        # This allows us to use bbox_inches=None for fixed image sizes
+        # Use 3 columns to ensure items wrap into multiple rows
+        ax.legend(
+            handles,
+            truncated_labels,
+            loc="upper center",
+            bbox_to_anchor=(0.5, -0.35),  # Lower in plot area, ~1.5 inches below chart
+            ncol=3,  # 3 columns ensures wrapping into multiple rows
+            fontsize=8,
+            framealpha=0.9,
+        )
+
+    # Explicitly set margins to ensure chart uses full width consistently
+    # Left margin (10%) accounts for y-axis labels (including formatted labels like "500.00M" or "$350.00")
+    # Right margin (5%) is minimal to maximize chart width
+    # Bottom margin (42.5%) accommodates legend positioned below the plot area (outside axes bounds) with ~1 inch of space below
+    # Top margin (8%) for title
+    # This ensures ALL charts have identical dimensions regardless of y-axis formatter
+    fig.subplots_adjust(left=0.10, right=0.95, top=0.92, bottom=0.425)
+
+    # Save chart to temporary file (use absolute path)
+    chart_filename = os.path.join(
+        output_dir, f"opik_chart_{chart_type}_{data['workspace']}.png"
+    )
+    chart_filename = os.path.abspath(chart_filename)
+
+    # Ensure output directory exists
+    chart_dir = os.path.dirname(chart_filename)
+    if chart_dir and not os.path.exists(chart_dir):
+        os.makedirs(chart_dir, exist_ok=True)
+
+    try:
+        # Use bbox_inches=None to preserve exact figure size (14x8 inches)
+        # Since legend is now inside figure bounds, we can use fixed dimensions
+        # This ensures ALL charts have identical dimensions (4200x2400 pixels at 300 DPI)
+        # regardless of y-axis label widths or content
+        plt.savefig(chart_filename, dpi=300, bbox_inches=None)
+        plt.close()
+
+        # Ensure file is fully written to disk using file system sync operations
+        # Retry loop to handle cases where file system hasn't fully flushed
+        max_retries = 10
+        retry_delay = 0.1
+        file_ready = False
+
+        for attempt in range(max_retries):
+            if os.path.exists(chart_filename):
+                try:
+                    # Try to open the file to ensure it's accessible
+                    with open(chart_filename, "rb") as f:
+                        # Force file system sync
+                        f.flush()
+                        os.fsync(f.fileno())
+
+                    # Verify file has content (size > 0)
+                    if os.path.getsize(chart_filename) > 0:
+                        # Verify file is readable
+                        if os.access(chart_filename, os.R_OK):
+                            file_ready = True
+                            break
+                except (OSError, IOError):
+                    # File may still be writing, wait and retry
+                    pass
+
+            if attempt < max_retries - 1:
+                time.sleep(retry_delay)
+
+        if not file_ready:
+            console.print(
+                f"[yellow]Warning: Chart file was not ready after {max_retries} attempts: {chart_filename}[/yellow]"
+            )
+            return None
+
+        return chart_filename
+    except Exception as e:
+        plt.close()
+        console.print(
+            f"[yellow]Warning: Could not save chart {chart_type}: {e}[/yellow]"
+        )
+        traceback.print_exc()
+        return None