Welcome to compact, schema-aware data assistant — think of it as an "EDA Copilot" powered by large language models (LLMs), built for analysts who want powerful, conversational control over their data.
This copilot allows users to explore tabular datasets using natural language. It provides an interactive command-line interface where prompts like:
"What are the top 5 features with missing values?"
or
"Simulate a 0.5% increase in interest rates and show the impact on loan status"
...are translated into real, pre-validated actions like SQL queries, pandas transformations, or simulations.
| Capability | Description |
|---|---|
| ✅ Schema Awareness | Understands and uses dataset schema for intelligent prompts and validation |
| 💬 Natural Language UI | Uses OpenAI’s API to interpret and map user queries to executable commands |
| 🧪 Secure Sandbox | All LLM calls are guarded — no dynamic exec() or eval() |
| 🔄 Interactive Shell | Run python your_tool.py to start an interactive session |
| 🧩 Fallback Support | Works even without an OpenAI key (manual commands) |
| 📊 EDA Utilities | Built-in functions for quick insights, summaries, and simulations |
| ✨ Portable | Cross-platform compatible, ≤ 400 logical lines |
data-copilot/ │ ├── your_tool.py # Main tool – single self-contained file ├── requirements.txt # Exact dependencies ├── .env.example # Template for API keys ├── demo.gif # [Optional] 2-min demo screencast ├── assets/ │ ├── banner.gif # Copilot banner / visual │ └── screenshots.png # Additional UI or CLI visuals └── README.md
"See it in action!"
Live demo screencast showing two interactions — schema understanding and a portfolio simulation.
- Built a lightweight command registry to ensure LLM safety and controlled execution
- Provided fallback CLI commands to allow full functionality without an API key
- Focused on modular actions (EDA summary, column stats, simulations) to encourage reusability
- Kept it under 400 lines, with clean docstrings and in-code comments
- Add Pinecone / Weaviate for vector search over schemas
- Embed streamlit or Gradio UI front-end for a visual interface
- Expand to support multi-table joins and basic forecasting capabilities
git clone https://github.com/yourusername/data-copilot.git
cd data-copilot
pip install -r requirements.txt
### Step 2: Add Your API Key
bash
Copy
Edit
cp .env.example .env
# Open .env and add your OpenAI API key
Step 3: Launch the Copilot
bash
Copy
Edit
python your_tool.py
🔒 Safety First
No eval(), exec(), or shell command execution.
Only whitelisted functions are allowed to run via the LLM interface.
---
✅ **To finish setup**:
- Replace image paths like `assets/banner.gif`, `assets/demo.gif`, or `assets/screenshots.png` with your actual files.
- Change the GitHub repo URL (`https://github.com/yourusername/...`) to your own.
- You can also embed the GIF using raw GitHub CDN links or use an external host like Imgur if preferred.
Let me know if you'd like a `requirements.txt`, `.env.example`, or help writing unit tests next!