vaibify 0.0.0

Vibe boldly. Verify everything.

Vibe Boldly. Verify Everything.

Overview

vaibify is based on the premise that AI-assisted research projects must be decomposable, verifiable, and containable.

vaibify creates secure, containerized environments for AI-assisted data science that can be accessed through a GUI and is fully scriptable. While it fully embraces agentic AI code development, vaibify also recognizes that a human must verify all results. vaibify solves these challanges by configuring Docker containers that prevent AI agents from harming your sensitive data, povides a GUI that supports terminal window(s) for running agents like Claude Code, and includes "viewing windows" that allow users to inspect resutls (data files, figures, animations). "Workflow" mode decomposes projects into automatic vs. interactive steps, and verifies the output via unit tests, dependency checks, and user validation. vaibify also seamlessly integrates with external resources like GitHub, Overleaf, and Zenodo so you can easily write reports, manage your work with version control, and archive your data. vaibify allows you to vibe code with confidence: your host machine stays safe while the agents develop code and build your analysis pipeline — all with minimal IDE interaction — enabling you to focus on vetting the results via visual inspection.

To get started, follow the Installation Guide and then the Quick Start.

Why Vaibify?

Data scientists increasingly rely on AI coding agents to build and iterate on analysis pipelines. But running AI-generated code raises real concerns:

• Safety — AI agents need broad permissions to be effective, but broad permissions on your host are dangerous. Vaibify runs everything inside an isolated Docker container with no access to your host filesystem, network, or credentials beyond what you explicitly grant.
• Reproducibility — Vaibify tracks provenance (SHA-256 hashes of every input and output), archives results to Zenodo with a DOI, syncs figures to Overleaf, and generates GitHub Actions workflows so anyone can reproduce your pipeline.
• Iteration — Decompose your project into steps, run them in parallel, inspect the outputs in the workflow viewer GUI, and re-run individual steps until you're satisfied.
• Generality — Vaibify is not tied to any specific domain. Configure your repositories, packages, languages (Python, R, Julia), and secrets in a single YAML file. Templates get you started fast.

Features

Vaibify provides a complete workflow for containerized scientific computing:

Container Management — Build, start, stop, and connect to Docker environments defined by a single vaibify.yml configuration file. Multiple projects can run simultaneously, each with its own container, image, and workspace volume. Target any project from any directory with --project/-p. Clone and install repositories, system packages, and Python/R/Julia dependencies automatically.

Pipeline Execution — Define multi-step workflows in workflow.json with data commands, plot commands, and test commands. Run individual steps or the full pipeline with one click in the browser-based GUI.

Workflow Viewer — A browser-based GUI for managing pipelines, viewing figures, monitoring resources, and running terminal sessions inside the container.

Security — No Docker socket inside the container, unprivileged user with gosu, ephemeral secrets mounted as mode-600 temp files, optional network isolation, and a built-in security audit (vaibify verify).

Reproducibility — Provenance tracking, Zenodo archival with DOI assignment, Overleaf figure sync, LaTeX annotation generation, and GitHub Actions workflow generation.

Templates — Two project templates ship with Vaibify: sandbox (no workflow, for exploration) and workflow (pipeline steps for reproducible analysis).

Quick Start

pip install vaibify
vaibify init --template workflow
vaibify build
vaibify start --gui

CLI Commands

vaibify init [--template NAME]          Create a project from a template
vaibify setup                           Launch the setup wizard GUI
vaibify build [-p NAME] [--no-cache]    Build the Docker image
vaibify start [-p NAME] [--gui]         Start the container
vaibify stop [-p NAME]                  Stop the container
vaibify status [-p NAME]                Show environment status
vaibify connect [-p NAME]               Shell into the container
vaibify verify [-p NAME]                Run the isolation security audit
vaibify gui [-p NAME]                   Launch the workflow viewer
vaibify push [-p NAME] <src> <dest>     Copy files into the container
vaibify pull [-p NAME] <src> <dest>     Copy files out of the container
vaibify config [edit|export|import]
vaibify publish [archive|workflow]

The shorthand vaib is also available.

Resources

The docs/ directory contains the full Sphinx documentation, also available online. The templates/ directory contains the project templates. The tests/ directory contains the pytest test suite.

Code Integrity

The Vaibify team maintains code integrity through automatic checks at every pull request. Unit tests run across all permutations of Ubuntu 22.04/24.04, macOS 15/26, and Python 3.9 through 3.14. Tests that require a running Docker daemon are excluded from CI and run locally. Documentation is rebuilt and deployed automatically on every merge to main.

Community

Vaibify is a community project. We welcome pull requests — please issue them to the main branch. See the Contributor's Guide for style conventions and testing instructions.

If you have questions or are running into issues, post to a Discussion.

If you believe you have encountered a bug, please raise an issue using the Issues tab.

Requirements

• Python 3.9+
• Docker (or Colima on macOS)
• macOS or Linux

License

MIT

© 2025 Rory Barnes.