Skip to main content

​
Overview

Sphinx offers multiple modes of interaction to suit different workflows and security requirements. Each mode provides a different amount of autonomy and control to the agent, allowing you to choose the approach that best fits your task.

Agent Mode

Sphinx works autonomously, executing code and iterating until the task is complete. Best for routine data science workflows where you trust Sphinx to make decisions.

Plan Mode

Sphinx creates a detailed plan before executing, giving you the opportunity to review and modify the approach. Ideal for complex analyses where you want strategic oversight.

Safe Mode

Every code execution requires your explicit approval. Perfect for sensitive operations or when you want to learn by reviewing each step.

Streamlit

Convert your notebook analysis into an interactive Streamlit application. Great for sharing insights with stakeholders through dashboards.

​
Choosing the Right Mode

ModeAutonomyBest For
AgentHighRoutine tasks, trusted workflows, rapid iteration
PlanMediumComplex analyses, strategic planning, multi-step tasks
SafeMinimalSensitive data, learning, careful review
Streamlit—Converting analyses to interactive apps

​
How to Switch Modes

You can switch between modes at any time using the mode selector in the chat input area. Click the mode icon (shown next to the input field) to see all available options and select your preferred mode.
Start with Agent Mode for most tasks. Switch to Plan Mode when tackling complex multi-step analyses, or Safe Mode when working with sensitive data or unfamiliar operations.