"""Reusable UI components for the MSEB Gradio Leaderboard."""
from __future__ import annotations
import gradio as gr
import pandas as pd
from src.config import TaskSpec
def create_header() -> gr.HTML:
"""Renders the top banner, title, description, and benchmark statistics."""
header_html = """
"""
return gr.HTML(header_html)
def create_search_bar(
placeholder: str = "🔍 Search models by name (e.g., 'gemini', 'flash', 'asr=truth')...",
label: str = "Search Models",
) -> gr.Textbox:
"""Creates a text input component for dynamic model name filtering."""
return gr.Textbox(
show_label=False,
placeholder=placeholder,
label=label,
elem_classes=["search-input"],
interactive=True,
)
def create_column_selector(
choices: list[str],
default_selected: list[str],
label: str = "Select Columns to Display:",
show_label: bool = True,
) -> gr.CheckboxGroup:
"""Creates an interactive checkbox group for toggling visible columns."""
return gr.CheckboxGroup(
choices=choices,
value=default_selected,
label=label,
show_label=show_label,
elem_classes=["column-selector-group"],
interactive=True,
)
def create_leaderboard_table(
df: pd.DataFrame,
) -> gr.DataFrame:
"""Creates a Gradio DataFrame configured for tabular leaderboard presentation.
Supports clickable Markdown model links and formatted numeric columns.
Args:
df: DataFrame containing leaderboard data.
Returns:
gr.DataFrame configured for display.
"""
return gr.DataFrame(
value=df,
interactive=False,
datatype="markdown",
elem_classes=["leaderboard-table"],
max_height=768,
wrap=True,
)
def create_task_header(
task_spec: TaskSpec,
dataset_count: int = 0,
) -> gr.HTML:
"""Renders task description and primary metric info callout banner."""
symbol = task_spec.metric_direction.symbol
direction_desc = (
"Higher is better"
if task_spec.metric_direction.name == "HIGHER_IS_BETTER"
else "Lower is better"
)
unit_desc = (
"Percentage [0–100%]"
if task_spec.default_unit == "percentage"
else "Raw distance metric"
)
html = f"""
{task_spec.display_name} ({symbol} {direction_desc})
{task_spec.description} •
Primary Metric: {task_spec.primary_metric} ({unit_desc}) •
Evaluated Datasets: {dataset_count}
"""
return gr.HTML(html)
def create_inactive_task_notice(task_spec: TaskSpec) -> gr.HTML:
"""Renders a notice for tasks with no evaluation records in the active results catalog."""
html = f"""
⚠️ Note: No evaluated models are currently available in
results/ for {task_spec.display_name}.
Showing the canonical column schema.
"""
return gr.HTML(html)