Using AI
What does AI assistance include?
Notebook.link integrates jupyterlite-ai, an extension designed for Jupyterlab based applications, that enables AI-powered code completions and chat directly in the browser using a specified provider.
JupyterLite-ai provides access to tools tailored for the Jupyter environment, such as:
- Notebook operations (creating files, adding/editing code, executing cells, etc.)
- Kernel execution (running code snippets without creating a new notebook)
- Jupyterlab commands (core and extension commands), allowing restrictions on certain actions—such as code execution—for security purposes.
note
Both the chat and code completion features require prior configuration. See Configure AI-addon for setup instructions.
JupyterLite AI settings
JupyterLite-ai settings are available via the dedicated settings panel as shown in the following screencast:
- Provider Management:
- View and configure available providers
- Set a default provider
- Configure MCP servers
- Behavior & Features:
- Enable context-aware responses
- Allow code execution capabilities
- Permit AI to use tools (built-in tools are related to the use of commands) -UI/UX Adjustments:
- Use Shift+Enter to send messages (Enter creates a new line)
- Display token usage in the chat toolbar
- Security & Commands:
- Add/remove commands requiring approval
- Prompt Customization:
- Modify the System Prompt (defines AI behavior and response style)
- Adjust the Completion System Prompt (guides AI-generated code)
- Diff View Options:
- Show diffs when AI modifies cell content or file content
- Choose a diff display mogde: split or unified
Chat
Once the AI add-on is configured, follow these steps:
- Load your project.
- Click the chat icon in the left sidebar to open AI chat panel.
- First use: The default provider is selected automatically.
- Subsequent uses: The last-used provider is retained (unless the browser cache is cleared).
- Switching providers: Use JupyterLite AI settings panel.
The video below demonstrates how to create and execute a notebook using the chat feature:
Code completion
The following video showcases an example of AI-assisted code completion: