Skip to content

Data Science & Engineering Components Guide

PyDataUI provides first-class native components built specifically for data professionals. Rather than wiring together complex frontend widgets, you can pass Python data structures, pandas DataFrames, and Plotly charts directly to PyDataUI components.

For complete page layouts, see Data Team Page Templates, which compose these components into analyst, data-quality, pipeline-operations, and ML-registry pages.


Table of Contents


For Data Analysts

DataFrameTable

Renders a pandas DataFrame as a scrollable, styled table with shadcn design patterns.

from pydataui.components.data_science import DataFrameTable
import pandas as pd

df = pd.read_csv("sales_data.csv")

# Render top 50 rows
DataFrameTable(df=df, max_rows=50)

Props: - df: pandas.DataFrame instance. - max_rows: Maximum rows to render (default: 100).

Rendered UI Preview
4 of 12,450 rows
order_id customer region amount status
#ORD-9021 Acme Corp North America $14,250.00 Completed
#ORD-9022 Starlight Media Europe West $8,920.50 Completed
#ORD-9023 BioGen Health Asia Pacific $31,400.00 Processing
#ORD-9024 Apex Logistics North America $4,110.00 Completed

DataTable (Universal Sniffing & In-Browser Interactivity)

Renders any dataset with automatic schema inference, instant client-side search filtering, interactive column sorting, and pagination.

from pydataui import DataTable

# Pass Polars, Pandas, DuckDB, Arrow, Pydantic, or Dicts directly:
DataTable(
    data=df,              # Any DataFrame, Relation, Arrow Table, or list of objects
    searchable=True,      # In-browser instant text filter
    sortable=True,        # Clickable column sorting (numeric & alphabetic)
    paginated=True,       # Client-side pagination
    page_size=10
)

Supported Inputs: - Polars DataFrames & LazyFrames (df.to_dicts()) - Pandas DataFrames & Series (df.to_dict('records')) - DuckDB PyRelations (conn.sql(...)) - PyArrow Tables & RecordBatches (table.to_pylist()) - Pydantic Models & Dataclasses ([Item(...) for ...]) - Columnar Dicts ({"col": [1, 2]})


PlotlyChart

Embeds interactive Plotly charts with auto-resizing and responsive layouts.

from pydataui.components.data_science import PlotlyChart
import plotly.express as px

fig = px.bar(
    df, x="quarter", y="revenue", color="region",
    title="Quarterly Regional Revenue",
    barmode="group"
)

PlotlyChart(figure=fig, height=450)

Props: - figure: Plotly Figure instance. - height: Pixel height of the rendered chart (default: 400).


MetricCard

Displays key business performance indicators (KPIs) with value formatting, delta indicators, and icons.

from pydataui.components.data_science import MetricCard

MetricCard(
    title="Monthly Active Users",
    value=284500,
    format=",.0f",             # Formats as "284,500"
    delta="+12.4% vs last mo",
    delta_type="increase",      # 'increase' (green) | 'decrease' (red) | 'neutral'
    icon="👥",
    description="Calculated based on unique 30-day token events"
)
Rendered UI Preview
Monthly Active Users 👥
284,500 +12.4% vs last mo

Calculated based on unique 30-day token events


For AI & ML Engineers

ModelMetrics

Grid of cards summarizing model evaluation metrics.

from pydataui.components.data_science import ModelMetrics

ModelMetrics(
    metrics={
        "Accuracy": 0.9542,
        "Precision": 0.9410,
        "Recall": 0.9680,
        "ROC-AUC": 0.9892
    }
)
Rendered UI Preview
Accuracy
95.42%
+1.2% benchmark
Precision
94.10%
High Confidence
Recall
96.80%
Minimal FNs
ROC-AUC
0.9892
Near-Optimal

ConfusionMatrix

Heat-mapped visual table representing model prediction accuracy across target classes.

from pydataui.components.data_science import ConfusionMatrix

ConfusionMatrix(
    matrix=[
        [1450, 42],    # [True Negative, False Positive]
        [28, 1380]     # [False Negative, True Positive]
    ],
    labels=["Legitimate", "Fraud"]
)
Rendered UI Preview
Confusion Matrix (Actual vs Predicted)
Pred: Legitimate
Pred: Fraud
Act: Legitimate
1,450
TN (97.2%)
42
FP (2.8%)
Act: Fraud
28
FN (2.0%)
1,380
TP (98.0%)

FeatureImportance

Horizontal relative-importance bar chart for model explainability (e.g. SHAP values, Random Forest Gini importance).

from pydataui.components.data_science import FeatureImportance

FeatureImportance(
    features={
        "transaction_amount_ratio": 0.385,
        "login_ip_entropy": 0.241,
        "card_age_days": 0.182,
        "failed_attempts_last_hour": 0.125,
        "is_foreign_country": 0.067
    },
    title="XGBoost Feature Importance"
)
Rendered UI Preview
XGBoost Feature Importance
transaction_amount_ratio 38.5%
login_ip_entropy 24.1%
card_age_days 18.2%
failed_attempts_last_hour 12.5%
is_foreign_country 6.7%

For Data Engineers

PipelineStatus

Real-time step tracker for ETL/ELT pipelines, Spark jobs, and DAGs.

from pydataui.components.data_science import PipelineStatus

PipelineStatus(
    steps=[
        {"name": "1. Extract from S3 Buckets", "status": "success", "message": "2.4GB in 12s"},
        {"name": "2. Data Cleansing & Validation", "status": "success", "message": "0 schema violations"},
        {"name": "3. PySpark Aggregations", "status": "running", "message": "Stage 4 of 8 active"},
        {"name": "4. Load to ClickHouse Data Mart", "status": "pending", "message": "Awaiting stage 3"},
        {"name": "5. Refresh Metabase Dashboards", "status": "pending", "message": "Awaiting stage 4"},
    ]
)
Status types: 'success' (green), 'running' (animated blue pulse), 'pending' (gray), 'failed' (red), 'skipped' (yellow).

Rendered UI Preview
✓
1. Extract from S3 Buckets
2.4GB in 12s
08:00:12
✓
2. Data Cleansing & Validation
0 schema violations
08:01:45
●
3. PySpark Aggregations
Stage 4 of 8 active (16 executors)
Running
○
4. Load to ClickHouse Data Mart
Awaiting stage 3
Pending

DataTimeline

Vertical activity stream and audit log for pipeline triggers and system events.

from pydataui.components.data_science import DataTimeline

DataTimeline(
    events=[
        {"time": "08:00:00", "label": "Scheduled Airflow DAG run triggered", "status": "info"},
        {"time": "08:01:12", "label": "S3 raw extract completed (14,200 partitions)", "status": "success"},
        {"time": "08:04:30", "label": "Great Expectations test suite passed", "status": "success"},
        {"time": "08:06:45", "label": "Cluster scaled from 4 to 16 compute nodes", "status": "info"},
    ]
)

SchemaViewer

Interactive table inspecting dataset columns, data types, null counts, and unique value counts.

from pydataui.components.data_science import SchemaViewer

# Option 1: Pass pandas DataFrame directly
SchemaViewer(df=my_dataframe)

# Option 2: Pass explicit schema dictionary
SchemaViewer(
    schema={
        "user_id": "int64",
        "email": "varchar(255)",
        "signup_ts": "datetime64[ns]",
        "lifetime_spend": "float64"
    }
)

SQLEditor

SQL query console with query parameter binding and an inline Run action button.

from pydataui.components.data_science import SQLEditor
from pydataui import State

class QueryState(State):
    query_text: str = "SELECT * FROM transactions LIMIT 100;"

    def run_query(self):
        # Execute query against database
        ...

SQLEditor(
    bind=QueryState.query_text,
    on_run=QueryState.run_query,
    rows=10,
    placeholder="Enter your SQL query here..."
)

FileDownload & DataUpload

In-browser data interchange components.

from pydataui.components.data_science import FileDownload, DataUpload

# Export CSV button
FileDownload(
    filename="aggregated_report.csv",
    content="date,metric\n2024-01-01,100\n",
    label="Download CSV Export"
)

# Upload dataset area
DataUpload(
    accept=".csv,.parquet,.json",
    on_upload="/api/data/upload"
)

JSONViewer & CodeBlock

Pretty-printed technical data displays.

from pydataui.components.data_science import JSONViewer, CodeBlock

# Collapsible formatted JSON
JSONViewer(
    data={"model": "Llama-3-70B", "quantization": "4-bit", "context_window": 8192}
)

# Syntax-aware code block with one-click copy button
CodeBlock(
    code="""from pyspark.sql import SparkSession
spark = SparkSession.builder.appName('Ingest').getOrCreate()
df = spark.read.parquet('s3://lake/raw/*')
""",
    language="python"
)