flowboard 0.1.2

Production-grade semantic analytics framework. Power BI-style modeling with DuckDB's sub-millisecond query performance.

Flowboard

Production-Grade Semantic Analytics Framework

Flowboard is a lightweight, intent-driven analytics engine that brings Power BI-style semantic modeling to Python. Build multi-table data models, query with natural intent, and visualize results—all with blazing-fast in-memory execution powered by DuckDB.

Why Flowboard?

• Semantic First: Define dimensions and measures once, query with intent
• DuckDB-Powered: Sub-millisecond analytics on datasets up to available RAM
• Multi-Table: Built-in relationship management across tables
• Zero Dependencies Hell: Minimal overhead, fast installation
• Production Ready: Designed for embedded analytics and data apps

---

Installation

pip install flowboard

Requires Python 3.8+

---

Quick Start: 60-Second Example

import flowboard as fb

# 1. Load data
sales_table = fb.load_csv('sales.csv')

# 2. Define semantic model
model = fb.SemanticModel()
model.add_table(
    'sales',
    dimensions=['date', 'region', 'product'],
    measures={
        'revenue': 'SUM(amount)',
        'profit': 'SUM(amount - cost)',
        'units': 'COUNT(*)'
    }
)

# 3. Query with intent
result = fb.query("revenue by region", model)

# 4. Visualize
chart = fb.chart(result)
chart.show()

---

Core Features

📊 Multi-Format Data Loading

fb.load_csv('data.csv')      # CSV auto-detection
fb.load_parquet('data.parquet')  # Columnar format
fb.load_xlsx('data.xlsx')    # Excel spreadsheets

🎯 Intent-Driven Queries

fb.query("revenue by month", model)
fb.query("profit by region", model)
fb.query("units by product", model)

🔗 Relationship Management

model.add_relationship('sales', 'customer_id', 'customers', 'id')

📈 One-Line Visualization

chart = fb.chart(result)  # Returns interactive Plotly figure
chart.show()

---

Industry Use Cases

E-Commerce Analytics

model.add_table('orders', 
    dimensions=['date', 'category', 'region'],
    measures={'gmv': 'SUM(total)', 'orders': 'COUNT(*)'}
)
result = fb.query("gmv by category", model)

SaaS Metrics Dashboard

model.add_table('events',
    dimensions=['event_type', 'cohort', 'date'],
    measures={'dau': 'COUNT(DISTINCT user_id)', 'sessions': 'COUNT(*)'}
)
result = fb.query("dau by cohort", model)

Financial Reporting

model.add_table('transactions',
    dimensions=['account', 'quarter'],
    measures={'revenue': 'SUM(amount)', 'margin': 'SUM(profit)'}
)
result = fb.query("revenue by account", model)

---

Architecture

┌─────────────────────────────────────┐
│        Flowboard API Layer          │
│  (Semantic Model + Intent Query)    │
└──────────────┬──────────────────────┘
               │
┌──────────────▼──────────────────────┐
│     DuckDB Execution Engine         │
│    (Sub-ms Query Performance)       │
└──────────────┬──────────────────────┘
               │
┌──────────────▼──────────────────────┐
│    Data Loaders (CSV/Parquet/XLSX)  │
│         + Relationship Mgmt         │
└─────────────────────────────────────┘

---

API Reference

load_csv(path) → str

Load CSV file into DuckDB. Returns table name.

load_parquet(path) → str

Load Parquet file into DuckDB. Returns table name.

load_xlsx(path) → str

Load Excel file into DuckDB. Returns table name.

SemanticModel()

Define your data model.

Methods:

• add_table(name, dimensions, measures) — Register a table with semantic metadata
• add_relationship(t1, col1, t2, col2) — Define foreign key relationships

query(intent, model) → list[dict]

Execute semantic query. Format: "<measure> by <dimension>"

chart(result) → plotly.graph_objects.Figure

Generate interactive bar chart from query result.

---

Performance Characteristics

Dataset Size	Query Latency	Memory Usage
< 100MB	< 1ms	Minimal
< 1GB	1-10ms	1-2GB
< 10GB	10-100ms	5-15GB

Benchmarks on modern hardware (2023+)

---

Development

git clone https://github.com/gyanankur23/flowboard
cd flowboard
pip install -e .[dev]
pytest

---

License

MIT — See LICENSE file for details.

---

Contributing

Contributions welcome! Please open an issue or PR on GitHub.

For questions and discussions, reach out to: [e-mail]

---

Flowboard v0.1.1 | Built for data teams, by data builders.