InsightAI 🚀

A powerful open-source library that enables natural language conversations with your data using Large Language Models (LLMs). Transform complex data analysis into simple conversations - no coding required!

✨ Key Features

• 🗣️ Natural Language Interface: Ask questions about your data in plain English
• 🔌 Multiple Model Support: Works with OpenAI GPT models and Groq's high-speed inference
• 🧠 Smart Analysis: Automatic code generation, data cleaning, and ML model suggestions
• 🛠️ Error Recovery: Built-in debugging and error correction mechanisms
• 📊 Auto-Visualization: Generates charts and graphs automatically
• 💾 SQL Support: Native support for SQLite databases
• 📈 Report Generation: Create comprehensive analysis reports automatically
• 🔍 Data Quality Analysis: Identifies and fixes data quality issues
• ⚡ Real-time Processing: Streaming responses for immediate feedback

🚀 Quick Start

Installation

pip install insightai-core

Environment Setup

Set up your API keys:

# Required: OpenAI API Key
export OPENAI_API_KEY="your-openai-api-key"

# Required: Groq API Key (for faster inference)
export GROQ_API_KEY="your-groq-api-key"

Basic Usage

import pandas as pd
from insightai import InsightAI

# Load your data
df = pd.read_csv('your_data.csv')

# Initialize InsightAI
ai = InsightAI(df)

# Start asking questions!
ai.pd_agent_converse("What are the main trends in this data?")

💡 Usage Examples

1. Interactive Data Analysis

import pandas as pd
from insightai import InsightAI

# Load sales data
df = pd.read_csv('sales_data.csv')
ai = InsightAI(df)

# Interactive mode - ask multiple questions
ai.pd_agent_converse()
# Now you can ask: "Show me monthly revenue trends"
# Or: "Which product category has the highest profit margin?"

2. Single Question Analysis

# Ask a specific question
ai = InsightAI(df)
ai.pd_agent_converse("What is the correlation between price and customer rating?")

3. SQL Database Analysis

# Analyze SQLite database
ai = InsightAI(db_path='customer_database.db')
ai.pd_agent_converse("Find the top 10 customers by total purchase amount")

4. Automated Report Generation

# Generate comprehensive analysis report
ai = InsightAI(df, generate_report=True, report_questions=5)
ai.pd_agent_converse()  # Generates a full report automatically

5. Data Cleaning and ML Suggestions

# Get data cleaning recommendations and ML model suggestions
ai = InsightAI(df)
ai.pd_agent_converse("Clean this dataset and suggest appropriate machine learning models")

🔧 Advanced Configuration

Constructor Parameters

InsightAI(
    df=None,                    # pandas DataFrame
    db_path=None,              # Path to SQLite database
    max_conversations=4,        # Conversation memory length
    debug=False,               # Enable debug mode
    exploratory=True,          # Enable exploratory analysis
    df_ontology=False,         # Enable data ontology support
    generate_report=True,      # Auto-generate reports
    report_questions=5         # Number of questions for reports
)

Custom Model Configuration

Create LLM_CONFIG.json in your working directory:

[
  {
    "agent": "Code Generator",
    "details": {
      "model": "gpt-4o",
      "provider": "openai",
      "max_tokens": 4000,
      "temperature": 0
    }
  },
  {
    "agent": "Planner",
    "details": {
      "model": "llama-3.3-70b-versatile",
      "provider": "groq",
      "max_tokens": 2000,
      "temperature": 0.1
    }
  }
]

Custom Prompts

Create PROMPT_TEMPLATES.json to customize agent behavior:

{
  "planner_system": "You are a data analysis expert...",
  "code_generator_system_df": "You are an AI data analyst..."
}

🎯 What You Can Ask

Data Exploration

• "What does this dataset contain?"
• "Show me the distribution of values in each column"
• "Are there any missing values or outliers?"

Statistical Analysis

• "What's the correlation between sales and marketing spend?"
• "Perform a statistical summary of the numerical columns"
• "Which factors most influence customer satisfaction?"

Visualizations

• "Create a bar chart of revenue by product category"
• "Plot the trend of monthly sales over time"
• "Show me a correlation heatmap of all numerical variables"

Data Cleaning

• "Clean this dataset and prepare it for machine learning"
• "Handle missing values and suggest the best approach"
• "Identify and fix data quality issues"

Machine Learning

• "What machine learning models would work best for this data?"
• "Prepare this data for predictive modeling"
• "Suggest features for predicting customer churn"

Business Intelligence

• "Generate a comprehensive analysis report"
• "What are the key business insights from this data?"
• "Create an executive summary of the findings"

📊 Output Examples

Automated Visualizations

InsightAI automatically saves visualizations to the visualization/ folder:

• Bar charts, line plots, scatter plots
• Correlation heatmaps
• Distribution plots
• Custom business charts

Analysis Reports

Generate professional markdown reports including:

• Executive summary
• Dataset overview
• Key findings and insights
• Recommendations
• Supporting visualizations

Code Generation

View the actual Python code generated for your analysis:

# Example generated code
import pandas as pd
import matplotlib.pyplot as plt

# Calculate monthly revenue trends
monthly_revenue = df.groupby('month')['revenue'].sum()
plt.figure(figsize=(10, 6))
plt.plot(monthly_revenue.index, monthly_revenue.values)
plt.title('Monthly Revenue Trends')
plt.savefig('visualization/monthly_revenue_trends.png')
plt.show()

🏗️ Architecture

InsightAI uses a multi-agent architecture with specialized AI agents:

• Expert Selector: Chooses the right agent for your task
• Data Analyst: Performs statistical analysis and visualizations
• SQL Analyst: Handles database queries and operations
• Data Cleaning Expert: Identifies and fixes data quality issues
• Code Generator: Creates Python code for your analysis
• Error Corrector: Debugs and fixes code issues automatically
• Report Generator: Creates comprehensive analysis reports

📈 Supported Models

OpenAI Models

• GPT-4o, GPT-4o-mini
• GPT-4 Turbo
• O1 series models

Groq Models (High-Speed Inference)

• Llama 3.3 70B
• Llama 3.1 8B
• Mixtral 8x7B
• Gemma 2 9B

📝 Logging and Cost Tracking

All interactions are automatically logged with detailed cost tracking:

{
  "chain_id": "1234567890",
  "agent": "Code Generator",
  "model": "gpt-4o-mini",
  "tokens_used": 1500,
  "cost": 0.03,
  "duration": "2.3s"
}

View logs in: insightai_consolidated_log.json

🔒 Security Features

• Input sanitization and validation
• Code execution sandboxing
• Blacklisted dangerous operations
• Rate limiting and error handling

🎓 Examples and Tutorials

E-commerce Analysis

# Analyze online store data
df = pd.read_csv('ecommerce_data.csv')
ai = InsightAI(df)
ai.pd_agent_converse("Which products have the highest return rate and why?")

Financial Data Analysis

# Stock market analysis
ai = InsightAI()
ai.pd_agent_converse("Download Apple stock data for 2024 and analyze the trends")

Healthcare Data

# Patient data analysis (anonymized)
df = pd.read_csv('patient_outcomes.csv')
ai = InsightAI(df)
ai.pd_agent_converse("What factors correlate with better patient outcomes?")

🛠️ Development Setup

git clone https://github.com/LeoRigasaki/InSightAI.git
cd InsightAI
pip install -e ".[dev]"

🆕 Version 0.5.0 Release Notes

✨ New Features

• Dynamic API Key Management: Only requires API keys for providers you actually use
• Flexible Provider Support: Mix and match OpenAI, Groq, and Gemini models freely
• Cost Optimization: Reduced overhead by eliminating unused API dependencies

🔧 Improvements

• Smarter LLM configuration parsing
• Better error messages for missing API keys
• Enhanced provider validation

🐛 Bug Fixes

• Fixed requirement for all API keys even when not needed
• Improved initialization error handling

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

1. Fork the repository
2. Create a feature branch: git checkout -b feature-name
3. Commit changes: git commit -am 'Add feature'
4. Push to branch: git push origin feature-name
5. Submit a Pull Request

⚠️ Known Limitations

• Token limits vary by model (check your plan)
• Large datasets may require chunking
• Rate limiting depends on your API plan
• Complex visualizations may need manual adjustment

📄 License

MIT License - see LICENSE for details.

🙏 Acknowledgments

• Special thanks to pgalko for the original inspiration
• OpenAI for providing powerful language models
• Groq for high-performance inference capabilities
• The open-source community for continuous improvements

💬 Support

• 📧 Email: riorigasaki65@gmail.com
• 🐛 Issues: GitHub Issues
• 💡 Feature Requests: GitHub Discussions

---

Transform your data analysis workflow today with InsightAI - where natural language meets powerful analytics! 🚀