Natural Language CSV/Spreadsheet Analysis — ask questions about your data in plain English and get instant answers, charts, and insights.
DataPilot turns any CSV, Excel file, or SQL database into an AI-powered analytics workspace. No SQL. No Python. Just ask.
- Upload your data — CSV, TSV, Excel, or connect a live database
- Ask a question — "Top 10 products by revenue" or "Average churn by segment"
- Get instant answers — Tables, numbers, and auto-generated charts
| Mode | What happens | Requirements |
|---|---|---|
| ☁️ File Upload | Drag-and-drop CSV/Excel, query with natural language | Any device, free API key (~1 MB default upload limit) |
| 🔌 Database Connect | Connect PostgreSQL, MySQL, SQLite, or SQL Server and query live tables | Database connection string |
git clone https://github.com/YOUR_USERNAME/datapilot.git
cd datapilot
python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
uvicorn app.main:app --host 0.0.0.0 --port 8065Then open http://localhost:8065, paste your free Groq API key into the Settings panel, and start asking questions.
See docs/SETUP.md for detailed installation and Docker instructions.
| Landing Page | Workspace | Auto-Generated Chart |
|---|---|---|
 |
 |
 |
- Backend: FastAPI (Python)
- Frontend: Jinja2 templates + HTMX + Tailwind CSS
- AI Engine: Groq Llama 3.3 70B (with Together AI & OpenRouter fallbacks)
- Data Processing: Pandas + Plotly
- Security: RestrictedPython sandbox for code execution
- Database: SQLite by default, PostgreSQL via Docker Compose
Read the full Architecture Overview.
| Capability | Description |
|---|---|
| Auto Pandas Code | LLM generates Pandas code from your natural language question |
| Auto SQL | For connected databases, generates safe SQL instead |
| Smart Charts | Automatically recommends and renders bar, line, pie, scatter, and heatmap charts |
| Column Fuzzy Match | Fixes typos and case mismatches in column names automatically |
| Error Recovery | Detects common mistakes and retries with corrected prompts |
| Sandboxed Execution | All generated code runs in a RestrictedPython sandbox |
| CSV Export | Download any query result or the full dataset as CSV |
DataPilot uses a multi-provider LLM backend. You only need one free key.
Recommended: Groq — fast inference, generous free tier
Alternatives: Together AI, OpenRouter
See docs/API_KEYS.md for the full provider list and troubleshooting.
pytestWith coverage:
pytest --cov=app --cov-report=term-missingWe welcome contributions! See docs/CONTRIBUTING.md for guidelines.
Priority areas:
- Additional chart types and customization
- Support for more database engines (BigQuery, Snowflake)
- Query caching with Redis
- UI/UX polish and accessibility
Please read our Code of Conduct and Security Policy before contributing.
- FastAPI for the backend framework
- Groq for blazing-fast LLM inference
- Pandas for data manipulation
- Plotly for interactive chart rendering
Note: File uploads are limited to ~1 MB by default (Starlette/FastAPI default). For larger files, connect a database or deploy behind a reverse proxy with chunked uploads. The UI loads Tailwind CSS, HTMX, Plotly, and Google Fonts from public CDNs, so an internet connection is required for the frontend to render correctly.