Turn your command line into an intelligent data analysis studio.

Data-Pilot is an interactive, agentic workflow engine that brings the power of LLMs directly to your local datasets.

Powered by the openai-agents runtime and wrapped in a stunning terminal interface, Data-Pilot bridges the gap between conversational AI and rigorous code execution.

Meet Vanessa, your dedicated AI analyst. She doesn’t just chat - she plans, executes Python code, builds models, and delivers reproducible insights, all while keeping your data secure within a local sandbox.

Why Data-Pilot?

Stop copying code snippets from a browser. Data-Pilot provides an integrated environment where reasoning and execution happen side-by-side.

Beautiful Terminal UX

Experience a modern command-line interface powered by Rich. Watch the agent’s reasoning stream in real-time, view color-coded tool outputs, and manage your session with intuitive slash commands.

Enterprise-Grade Guardrails

Security by design. The agent operates exclusively within a ./root sandbox, ensuring that your system files remain untouched. Every action—from file reads to code execution—is strictly scoped to the project environment.

Batteries-Included Toolbelt

Data-Pilot comes pre-loaded with a robust suite of analytical tools:

• Automated EDA: Generate quality reports, correlation matrices, and schema overviews instantly.
• Modeling Pipelines: Run end-to-end baseline training (Random Forest, Logistic Regression) with a single prompt.
• Python Execution: The agent writes and runs arbitrary Python to solve complex logic problems on the fly.

Extensible Architecture

Built on a modular stack, Data-Pilot allows you to plug in new tools or “handoff” agents with just a few lines of configuration. Whether you are using Cerebras or OpenAI, the system adapts to your infrastructure.

The Workflow

1. Ingest: Drop your CSV, JSON, or Parquet files into the sandbox.
2. Prompt: Launch main.py and state your business objective.
3. Plan: Watch as the agent drafts a multi-step analysis roadmap.
4. Execute: The agent writes code, runs it, and interprets the stderr/stdout logs automatically.
5. Deliver: Receive a summarized report with artifacts, metrics, and charts saved to your output directory.