Jupyter Notebook Tool

A Python package providing atomic tools for langchain-based AI agents to manipulate Jupyter notebooks. Built on nbclient/nbformat, it enables programmatic notebook creation, loading, and manipulation.

Installation

You can install from this repo in development mode with pip install -e ./src/

Core Functionality

The system maintains notebook sessions using tokens. Each operation requires a valid token obtained from either:

• create_notebook(): Creates new empty notebook
• load_notebook(url): Loads notebook from URL or file path

Available Operations

With a valid token, the following operations are supported:

• list_cells(token): Get ordered list of cell IDs
• create_cell(token, source, cell_type, position): Add new cell
• update_cell(token, id, source): Modify cell content
• execute_cell(token, id): Run code cell and get output
• delete_cell(token, id): Remove cell from notebook
• get_notebook(token): Retrieve current notebook state

Error Handling

The system provides specific error types for common failure cases:

• InvalidTokenError
• CellNotFoundError
• CellTypeError
• NotebookLoadError
• KernelError

Worked Example

See docs/example.ipynb for an example of how to use the package.