Audit & Environment Capture

The floability audit command automatically discovers the software and data dependencies of a Jupyter notebook by executing it and recording what packages and files it uses. Instead of manually curating an environment.yml, you run the notebook once and Floability captures the exact environment.

Overview

Floability executes the notebook under system-call tracing, capturing every package loaded from site-packages and every data file opened during execution. It separates these into manager-side (notebook process) and worker-side (TaskVine worker) dependencies, then generates verified environment YAMLs pinned to the exact installed versions in your active environment.

Outputs

After a successful audit, the following files are written to the current directory:

File	Description
manager_environment.yml	Conda/pip environment for the notebook manager process
worker_environment.yml	Conda/pip environment for TaskVine workers
manager_data_dependencies.yml	Data files opened by the manager, with file sizes
worker_data_dependencies.yml	Data files opened by workers, with file sizes

These environment files can be used directly as the software/environment.yml in a backpack.

Basic usage

floability audit --notebook path/to/my_notebook.ipynb

Options

# Use a specific Jupyter kernel (name as shown by `jupyter kernelspec list`)
floability audit --notebook my_notebook.ipynb --kernel python3

# Connect the local worker to a named TaskVine manager
floability audit --notebook my_notebook.ipynb --manager-name my-manager

# Generate per-cell dependency breakdowns
floability audit --notebook my_notebook.ipynb --cell-level

Cell-level audit

The --cell-level flag produces an additional cell_level_dependencies.yml that lists the code and data dependencies for each notebook cell individually. This is useful for understanding which cells drive which dependencies and for scoping worker environments to only the cells that run remotely.

notebook_name: my_notebook.ipynb
cells:
  - cell_number: 1
    code_dependencies:
      - numpy==1.26.4
      - pandas==2.2.1
    data_dependencies: []
  - cell_number: 2
    code_dependencies:
      - matplotlib==3.8.4
    data_dependencies:
      - /home/user/data/input.csv

Requirements

• strace must be available on the system (Linux only).
• vine_worker (from ndcctools) must be installed and on PATH.
• The notebook must be fully executable in the current environment before auditing.