curl | bash for AI Agents: One-Command Install for pydantic-deep

TL;DR: The standard Python CLI tool install is 7 steps and 15+ minutes: Python version, venv, pip install, wrong extras, PATH confusion. pydantic-deep v0.3.6 fixes this with a one-command installer: curl -fsSL https://raw.githubusercontent.com/vstorm-co/pydantic-deep/main/install.sh | bash. The script auto-detects uv, installs it if missing, runs uv tool install "pydantic-deep[cli]", and verifies PATH. pydantic-deep update handles upgrades. Startup update notifications check PyPI with a 2-second timeout and 24-hour cache — they never block startup.

Here’s how you install most Python AI tools today:

1. Install Python. (Which version? 3.10? 3.11? 3.12? Check the README. The README says “Python 3.8+.” That’s not helpful.)
2. Create a virtual environment.
3. Activate it. (Remember every time.)
4. pip install tool-name — wait for dependencies.
5. Run the tool. Get ModuleNotFoundError: No module named 'textual'.
6. pip install tool-name[cli] — the extras were in a separate bracket.
7. Try again. Now it works.

Seven steps. Fifteen minutes minimum. For a first impression.

This is the standard Python CLI tool install experience, and it is a real barrier — for developers who work in other languages, for teams onboarding new people, for anyone who wants to try something quickly before deciding whether it’s worth their time.

We decided to fix this for pydantic-deep — the modular agent runtime for Python.

I’m Kacper Włodarczyk, co-founder of Vstorm — an Applied Agentic AI Engineering consultancy. We’ve deployed 30+ production AI agents. pydantic-deep is our open-source AI coding agent.

One Command

curl -fsSL https://raw.githubusercontent.com/vstorm-co/pydantic-deep/main/install.sh | bash

That’s the entire install. No Python version management. No virtual environments. No extras brackets.

What install.sh Actually Does

The install script is short and readable — inspect it before running (and you should, with any curl | bash). Here’s what it does:

Step 1: Detect uv

uv is a Python package manager from Astral that handles virtual environments, tool installs, and Python version management transparently. If you have uv, the script uses it immediately. If not, it installs uv via the official Astral installer.

if command -v uv &> /dev/null; then
    echo "uv found, using existing installation"
else
    echo "Installing uv..."
    curl -LsSf https://astral.sh/uv/install.sh | sh
fi

Step 2: Install pydantic-deep as a uv tool

uv tool install "pydantic-deep[cli]"

uv tool install creates an isolated environment for the tool and makes the binary available globally. No activation needed. The [cli] extras group includes everything required to run the full TUI — including textual.

Step 3: Verify and guide

The script runs pydantic-deep --version to verify success. If the binary isn’t in PATH (common on first uv install), it prints exactly what to add to your shell config — the correct line for bash or zsh — and explains why.

The textual Bug That Started This

Before v0.3.6, the standard install was pip install pydantic-deep. Then we started getting ModuleNotFoundError: No module named 'textual' from fresh installs.

The issue: textual was in the tui extras group, but most users didn’t specify any extras. We moved textual into the cli extras group — so pydantic-deep[cli] now includes everything needed for the full experience.

The install script handles this transparently.

Self-Update: pydantic-deep update

Updating is:

pydantic-deep update

This is brew upgrade pydantic-deep for AI agents. Under the hood:

• If uv is available: runs uv tool upgrade pydantic-deep
• If not: falls back to pip install --upgrade "pydantic-deep[cli]"

No manual version checking. No reading changelogs to decide whether to update.

Startup Update Notifications

Every time you start pydantic-deep, it quietly checks PyPI for a newer version:

pydantic-deep v0.3.6
Update available: v0.3.6 → v0.3.7  Run: pydantic-deep update

Implementation details:

• 2-second timeout — never blocks startup
• Results cached in ~/.pydantic-deep/update_check.json for 24 hours
• Hits https://pypi.org/pypi/pydantic-deep/json — PyPI’s public JSON API
• Implemented with urllib and threading — zero additional dependencies

Why uv?

We evaluated several options:

pipx — similar concept, good ecosystem support, but slower and requires its own separate install first.

Homebrew formula — great UX but requires maintaining a tap, and Python package updates don’t map cleanly to Homebrew’s lifecycle.

Native binary (PyInstaller/Nuitka) — eliminates the Python dependency but our tool uses dynamic imports that made compiled output brittle across platforms.

uv — fast, handles Python version management transparently, uv tool is designed exactly for this use case. The tradeoff: uv gets installed on the user’s system. But uv’s own install is excellent and most Python developers will find it useful beyond pydantic-deep.

The Bigger Picture

Developer experience is a feature. The time between “I want to try this” and “I’m running this” is a conversion metric. We want that gap to be 30 seconds, not 15 minutes.

The same philosophy applies to pydantic-deep update. You should never be stuck on an old version because updating is annoying.

Key Takeaways

• The standard Python CLI install experience is a barrier — venv management, PATH issues, wrong extras.
• install.sh handles everything: uv detection, uv install if needed, pydantic-deep[cli], PATH verification.
• pydantic-deep update is one command to stay current — uv tool upgrade or pip fallback.
• Startup notifications check PyPI passively with a 2-second timeout and 24-hour cache.
• textual is now properly in cli extras — the full TUI always works.

Try It

curl -fsSL https://raw.githubusercontent.com/vstorm-co/pydantic-deep/main/install.sh | bash

# Later, to update:
pydantic-deep update

GitHub: github.com/vstorm-co/pydantic-deep | OSS portal: oss.vstorm.co

What’s the worst install experience you’ve had with a Python AI/ML tool?

Vstorm builds production AI agent systems. More open-source tools at oss.vstorm.co.