Data Analysis and Visualization Agent

Intelligent data analysis system that converts natural language queries into SQL, executes them, generates insights, and creates visualizations.

Features:

• Natural Language to SQL: Ask questions in plain English, get accurate SQL queries
• Multi-Database Support: Works with Chinook, Can be extended to new databases easily
• Intelligent Visualizations: Auto-generates appropriate charts (bar, line, pie, scatter)
• Role-Based Security: Fine-grained table access control by user role
• Conversation Memory: Multi-turn conversations with context awareness
• Intent Classification: Distinguishes data queries from off-topic questions
• Self-Healing/Correction SQL: Automatic retry with error correction on query failures
• Observability: Langfuse integration for tracing and monitoring

Architecture:

We are using Langgraph framework to orchestrate the workflow for agents

LangGraph Workflow:

Key Components:

• Intent Classifier: Routes queries (data_query, follow_up, off_topic)
• Analysis Agent: Generates and validates SQL with structured outputs
• Visualization Agent: Recommends and creates charts based on data characteristics.
• Conversation Manager: Maintains session context for follow-ups
• Permission Manager: Enforces role-based table access

Demo:

Click to watch the demo video

🚀 Quick Start

Installation:

# Clone the repository
git clone https://github.com/Asad-Ismail/query-analysis-plot-agent.git
cd query-analysis-plot-agent

# Install dependencies
pip install -r requirements.txt

# Or use uv (faster and recommended)
bash uv_install.sh

Configuration:

Create a .env file:

OPENAI_API_KEY=your_api_key_here
OPENAI_API_BASE=https://api.openai.com/v1  # Optional custom endpoint
LANGFUSE_SECRET_KEY = langfuse_secret_here
LANGFUSE_PUBLIC_KEY = langfuse_key_here
LANGFUSE_BASE_URL = langfuse_base_url

Dataset Setup:

Download one or more of the test sql database from here

Place them in the root data/ directory

Usage:

CLI Mode

# Single query
python cli.py "Show top 5 artists by sales" --database chinook

# Interactive mode
python cli.py -i --database chinook

Web Interface

# Start backend
python app.py

# Open frontend
open index.html

Access at http://localhost:5000

Example Queries:

# Simple retrieval
"Show me top 3 artists by revenue"

# Aggregation
"How many customers do we have in total?"

# Filtering
"List all customers who live in Berlin"

"Who are our top 5 customers by total spending?"

Testing:

# Run functional tests
pytest -s -v tests/test_functional_chinook.py

Tools Used

• LangChain & LangGraph: Workflow orchestration
• OpenAI GPT-40-mini: LLM for SQL generation and insights
• Langfuse: Observability and tracing
• Pandas: Data manipulation
• Matplotlib: Visualizations
• Flask: Web backend
• SQLite: Database engine